Rising bond yields sparked a sell-off in AI shares this week. Whereas initially the AI buildout was largely financed by strong operating cash-flows of the internet giants and large cloud players; in recent quarters, even these big players are now burning cash in order to accelerate their AI buildouts.

The reason is simple—as we calculated last week; revenue generation from AI is already very strong and the IRR on the massive AI buildout over the past 2.5 years is very attractive. Financial commentators all take into account the massive AI capex that Meta and Google are making, but attribute zero benefit from AI to their growth acceleration in ad revenues. We think this is a very flawed assumption as we detailed below.

Over the past 2.5 years—since the start of the AI data center buildout—typically it cost around $38 billion to build one gigawatt of AI capacity. At current market rates, you can rent out that capacity at $60 billion per annum (based on SpaceX’s latest deals). So, anyone who made that investment over the last 2.5 years and now has capacity available is currently printing money with those GPUs. The payback on that data center is less than one year.

Even better, according to Nvidia’s data, those GPUs will run for six years or even longer. Nvidia shows that 6.5 years old A100s are currently being sold at 25% of the original purchase price still, while recent GPU versions are being sold at a premium:

However, with bond yields now spiking, this is putting pressure on the AI buildout as it’s partly also being financed with debt. Smaller fish such as the neoclouds, for example, are heavily financing their data center buildouts with leverage.

Over the past few weeks, France has been at the center of the storm in the bond markets, with spreads widening to similar levels to those seen during the Eurozone government debt crisis of 2012.

Investors who purchased French government bonds at negative rates over the 2019-2021 period might well have made the worst investment ever. They essentially purchased bonds which they thought were risk-free, they’re still paying the French government currently to hold their money, and they might never get their money back as left-wing political candidates are now talking about simply ‘cancelling the debt’.

Louis Gave from Gavekal explains the dire situation France is facing on the Macro Voices podcast:

“France has been running roughshod around all the various European rules, running budget deficits essentially of 5% of GDP. You could easily tell me “yeah, so what, other countries do so.” I think there are a few massive differences for France today which explain why France has essentially started what looks and feels like a typical emerging market debt crisis. You could say, well, the US is running budget deficits of 7% of GDP and that seems to be going well. Why should France hit the skids at five? There are a few things to consider.

The first is that France’s ability to increase taxes is essentially zero. France’s total tax receipts account for roughly 44% of GDP. In the US you’re at about 24–25%. So right now in France, we’re on the wrong side of the Laffer curve. If we try to meet these massive budget deficits through tax increases, we’ll actually be raising less money. That’s the first problem.

The second massive problem is that a majority of our debt is now owned by foreigners. I just spent the past couple of weeks on the road meeting with clients, and it was actually interesting. The first question most clients ask is: at what level of yield should I buy US government bonds? Are we there yet? Have yields risen enough in the US that bonds are now attractive? That’s the first question. And invariably, the second one almost in the same breath is: who’s stupid enough to own French bonds here? Who’s going to be the bag holder of this unfolding train wreck?

For a lot of people owning France—and especially French debt—has now become a career risk. If you buy French bonds and France ends up restructuring its debt or the euro blows up or whatever else, you look like an idiot because for the past couple of years it’s been all over the Wall Street Journal and this was one of the most predictable crises ever. So that’s the second issue.

But the third issue—I mentioned that France was going through some kind of emerging market debt crisis. The reality is, like an emerging market, France has issued a bunch of debt in a currency that we can’t print. We depend on other people to print the money to bind us over at this stage, and that’s the crux of the matter. Now you could say, “well, the ECB will, push comes to shove. At the end of the day, the ECB is run by a French lady, so push comes to shove, things will happen.” Which, as an aside, is perhaps one of the oddest things right now in the market: while France is on fire, the market is still pricing in three rate hikes from the ECB. That’s never going to happen. The ECB is never going to have three rate hikes.

But in any event, what we have today in France is a confluence of different crises. We have a fiscal crisis, we have an energy crisis, we have an economic crisis, we have a social crisis, and we have a political crisis all at once. And all this is occurring at a time when we have a president who’s in office but not in power. His opinion polls at this stage are down in the low teens and he has absolutely no authority whatsoever. And so with this lack of authority, riots are running wild. When you look at France, the bottom line is if we want to get out of our current economic bind, there’s only two choices. Either we have to move the retirement age to 70 years old, option one, which is extremely hard to do, or like in any good emerging market crisis, you have to devalue the currency by 50%. But within the euro, that’s not possible.

One of the very disappointing things of this French presidential election that is coming up in six months is you have lots of candidates but not a single candidate proposing either of these options, when economically you’re going to need to do one or the other and perhaps both. What you do have is a far-left candidate essentially saying look, we need to default on the debt, which, once a majority of your debt is held by foreigners, is actually not completely stupid. So for Italy, for example, it makes no sense for Italy to default on its debt because 80% of the debt is held by Italians. Why would you default on your own people?

Defaulting on foreigners—yeah, maybe that makes sense. But, if you go down this path, you’re essentially bankrupting the entire European insurance industry. France has issued more debt than the entire African continent. A lot of our debt is essentially held by European pension funds and European insurance companies. So the first thing you do if France goes down this path is essentially blow up the European insurance industry and probably the global one, because Japanese insurance companies, US insurance companies, a lot of people own French debt.

The second issue, we have three banks in France that are systemically important banks and that are huge in the derivatives markets. If all of a sudden French debt is redenominated, the balance sheets of a lot of these banks, they have liabilities in euros, all of a sudden their assets are redenominated in a French franc that loses 40%. That could create some serious mishap. Let’s not pretend that this would be easy and that it would not be deeply disturbing for global financial markets.

The reality is, if France defaults on its debt, restructures its debt, leaves the euro—pick any scenario—if any one of those things happens, it’s very hard to see the euro surviving as it is today because at that point the European financial industry, the banks, the insurance companies, basically all face catastrophic losses and every country will need to nationalize their own debt.”

So, this sounds pretty bad—the situation in France could spark a wider financial crisis due to its contagion effects. However, we’d make a few points here.

In hindsight, although there was substantial market turmoil at the time, the Euro debt crisis 1.0 was actually resolved pretty easily. On July 26, 2012, ECB President Mario Draghi spoke three magic words—“whatever it takes”. Basically, he signaled to markets that the ECB would purchase as many government bonds as necessary in order to keep borrowing yields under control. After this, markets calmed down and bond yields kept declining for another 8 years.

Obviously, if we get a similar solution now—with the ECB basically promising to purchase as much Eurozone government debt as necessary—the problems will come back at some stage in the future. Governments will keep running large deficits, government debt will keep building, and at some stage you run into serious macro consequences such as high inflation rates. Argentina, Turkey, and other emerging markets are well aware of this. In Europe, however, besides the Germans, most politicians are likely ignorant of this topic.

We’re less concerned about high bond yields in the US. If you look at long term history, bond yields have actually simply normalized in recent years. Historically, the US economy has also functioned well with bond yields at 5% and even higher.

Taking a contrarian view on the macro, more from a bottom-up perspective, we estimate that AI will be a hugely deflationary force in the world. What took a programmer two years to code, Claude can now do in a single day. All sorts of basic digital tasks can now be replaced with smaller, open-source LLMs at prices that are not much more expensive than electricity. At the same time, AI is driving huge productivity gains and advances in engineering, sciences, legal work, research, etc. We see both simple and advanced AI models making large contributions to the global economy over time, which should be highly deflationary and push interest rates down again.

Blackstone’s President during a recent presentation mentioned that US productivity growth has started to accelerate:

“Annual US productivity growth, which for a decade, you know, averaged 1.5%, is now 2.6% over the last two and a half years. It’s hard to say how much is AI-driven, but something is happening out there and we’re beginning to see it. Productivity has made a step function move upward.

The tip of the spear is the hyperscalers. Their revenue per employee has grown 65% over a 3.5-year period as they adopt these technologies. And of course, it’s showing up in margins. The EBITDA margins for the S&P and our portfolio companies are up 500 and 700 basis points over this four-year period of time. And if you talk to our CEOs, they would tell you the next 12 months are going to be very positive as well.”

When it comes to high government debt levels, Japan shows that nations can survive on extremely high levels of government debt. While many macro economists have been screaming for 30 years now that Japan is going to collapse, so far this hasn’t happened.

Of course, we wish that governments across the world would be running small surpluses and have virtually no debt at all, it would be much better, but we’re not seeing anything above that makes us want to sell our equity portfolio and panic.

Another large binary risk, however, is a potential energy crisis. According to news reports, the US has prepared a three day strike plan to finish the Iran conflict. Basically, the US has been monitoring all communications of the IRGC over the last six months and so they know exactly where all Iranian nodes are operating from. However, the risk is that the Iranians would successfully retaliate and take out energy infrastructure across the Gulf countries. A recent article in CBC highlighted the risks of this strategy:

Citrinowicz, who’s with Israel’s Institute for National Security Studies, said a third round of war could be the worst yet, based on what Iranian officials are telegraphing.

“They will attack the energy facilities in the Gulf. But more than that, they will cause significant damage to civilian infrastructure, communications, as well as cutting off the internet lines going under [the] Hormuz Strait,” he said.

He said there is even a possibility that Iran could push some of its massive land army into the fight with limited assaults against Gulf countries that house U.S. bases, such as Kuwait and Bahrain.

“We’ll see a regional war with Iran investing a lot in attacking the U.S. presence in the region to cause maximum economical damage to the U.S. international economy.”

Obviously, this is a substantial risk for the global economy and if Iran would successfully take out energy infrastructure across the Gulf, we expect a strong market sell-off. However, this is the reason that we typically invest in extremely strong companies and with no leverage in the portfolio. When you invest in the types of companies where if you look out five years or longer, and you can only imagine share prices being substantially higher regardless of all the possible scenarios when it comes to the macro environment, this allows you to go through short term market turmoil much easier.

What we typically do during times of market volatility, is simply go through the list of holdings in our portfolio and review valuations. Typically, we find valuations extremely attractive and that makes us want to buy more of those names. But, of course, for this strategy to be successful, the quality of your portfolio holdings has to be good.

For example, when it comes to investing in AI clouds, we hold a large position in Amazon but we don’t hold any CoreWeave. This turned out to be a good call in the current environment of rising interest rates and it allows our portfolio to navigate current macro turmoil much easier.

While CoreWeave has a massive AI backlog for a neocloud, all of this data center capex will have to be financed with debt. CoreWeave is operating on extremely high leverage and this makes the market nervous in the current environment.

However, it’s not all doom and gloom for CoreWeave. As we discussed in our opening, AI is actually generating attractive revenues already while AI capacity is still seeing huge shortages. This gives scope for CoreWeave to increase pricing on new contracts over time. Goldman thinks that also financing costs will remain under control:

“Going back to Finance 101, we see CoreWeave’s enterprise value as being directly correlated to the magnitude and durability of the spread between its unlevered IRR and WACC — and we think investors may be underestimating the magnitude of upward revisions to uIRR, and overestimating the magnitude of upward revisions to WACC:

We believe CoreWeave’s ARR per MW, and therefore unlevered IRR, is biased to the upside in the next 24 months because of tight supply, as well as our view that CoreWeave is differentiated from other neoclouds in its performance/cost metrics and its ability to execute on ramping new supply;

We estimate CoreWeave’s blended cost of debt could move 100-200bps but likely not more than that, ultimately putting a cap on WACC such that equity return profiles will still be attractive, based on feedback from our industry credit conversations. While unsecured debt trades at 13.5-15.0% YTW (Yield to Worst), we do not think that this would be representative of the company’s new cost of financing as CoreWeave’s financing structure is primarily asset-based.”

With CoreWeave now trading at 4x forward Sales, we do think there is an investment case here given the company’s strong growth profile and the attractive outlook in AI. However, we prefer to hold large positions in the big cloud platforms as we see these as much superior platforms to run enterprise cloud workloads on—neoclouds can’t handle most of these workloads, they largely provide just GPUs—and with a much lower risk profile, especially in the current environment.

Despite the macro turmoil, everything is still looking strong in semis. Deutsche notes that mature semis, for example, are seeing strong price increases:

“In partnership with dbDataInsights, we have been conducting a weekly scrape of chip pricing of mainly analog/discrete/MCU components sold into the distributor channel (>1,500 SKUs). During October, the trend is one of intensifying tightness with mature chip pricing trending upwards as shortage conditions spread from specific lagging-edge categories (where AI demand plays an outsized role) to a broad-based picture of tightness across mature process nodes, not helped by TSMC. Lead times in power discretes have reached 40-52 weeks, with a demand pickup also seen in SiC MOSFETs. Notably, onsemi announced a sweeping price increase of 10-20%, which took effect on October 10th. In analog, lead times at TI and ADI have extended in some parts up to 60-70 weeks. However, outside of power/analog, allocation notices have become more common in certain 32-bit MCU families (eg STM32 lead times extended to 40-52 weeks, Infineon, Renesas). Lastly, passives (notably high-end MLCCs) continue to see stretched lead times.”

Texas Instruments is seeing broad-based demand strength across the data center and industrial end markets—including aerospace and defense, energy infrastructure, test and measurement, and robotics:

“The data center market, it’s a market that we see strong secular growth. I don’t think it’s influenced by seasonality. The demand continues to be strong and growing. It’s just a bigger part of our business now. It was about 13% of revenue in Q3 or so. On top of it, I’m very excited about what we are seeing on the industrial market. We are seeing a very strong momentum in industrial across the sectors, but specifically in areas like aerospace and defense, energy infrastructure, test and measurement, the robotics sector. So this is where we see a combination of secular growth, but also sequential or cyclical recovery. So this is why we were seeing an above-seasonal forecast from TI. We also saw the automotive customers waking up, if you will, with higher demand in Q3, and I think that may continue into Q4. So again, the top 3 markets, industrial, automotive and data center, are the ones that are driving our guide into Q4.”

Semi shortages are even worse at the leading edge, which powers the AI workloads. Deutsche doesn’t expect supply to catch up with demand until 2029:

“The leading-edge logic and memory shortage remains acute. On the foundry side, we don’t expect supply to catch up with demand until 2029, and today lead times are heavily extended on advanced nodes (2-3nm), stretching to nearly 3 years at some extremes. On the memory side, there has been significant discussion since the summer around HBM de-speccing back down to 8-hi stacks to drive higher xPU volumes, but more recently there have been reports around execution issues tied to MR-MUF at SK Hynix with 12-hi stacks, which feels more intuitive. Either way, DRAM serving non-hyperscalers remains extremely tight, with the best supply alternatives coming from China (eg CXMT). As for server CPUs, Fusion reports that lead times of Intel SKUs sit at around 50 weeks. EUV has also become a bottleneck.”

There were concerns in the market around memory de-speccing, however, the long term trend is that advanced AI will continue to become more memory intensive. Micron had a nice chart on this topic at their tech symposium this week:

Overall, while we can’t forecast what’s going to happen in the Middle East, we do think there is a lot of cause for optimism when it comes to AI and tech in the coming years. We can see scenarios where there is substantial market turmoil in the coming months, however, we feel very comfortable with the quality and valuations of the businesses we own. So, we wouldn’t sell any of our holdings, although a partial hedge via futures contracts could be appropriate at some stage to reduce volatility.

While some macro fundamentals look dire—especially when it comes to government finances—we can see AI improving macro fundamentals over time by raising productivity growth and bringing down prices. This should result in higher GDP growth, lower inflation, lower interest rates, higher corporate margins, and even increased government efficiency (which can help lower high current deficits).

Personal AI Agents – Winners & Losers

Read more

Anthropic’s slowing growth rates created substantial anxiety in the market this week. For example, Reuters discussed a lower valuation for the company’s upcoming IPO:

“Some current Anthropic investors and prospective investors in its upcoming initial public offering are weighing a lower valuation ‌for the AI company after its latest financial disclosures, four people familiar with the matter said, amid broader concerns over the ballooning cost of financing AI.

Some investors who attended Anthropic’s investor day on Wednesday in San Francisco discussed a valuation of around $1.5 trillion, three of the people said, a significant step down from the roughly $2 trillion valuation ​that investors had previously expected for the November share sale.”

The reason is that over the past two months, sequential growth rates for Anthropic slowed to below 7% month-on-month. We can also see that OpenAI’s ARR growth accelerated to 16-19%:

However, we’d make a few points here. Firstly, there is seasonality. Most people take their annual holiday in August, and so one would expect enterprise token demand to be lower in that month. Despite this, both Anthropic and OpenAI still showed attractive growth.

Confirming the seasonality—we can see that GPU pricing weakened over August and recently started picking up again. Therefore, we expect Anthropic’s ARR growth to accelerate again in the coming months.

Secondly, September is the month that OpenAI released Astra. You would expect dollars to shift to Astra to try it out. For example, we did a large coding project with Astra last month. However, this month, we’ve already switched back to Claude. Astra is definitely a smart model but overall Opus 5.5 is the far more productive and better model to work with. And now Anthropic will be releasing the new Fable in the coming month, which no doubt will again even have stronger capabilities.

At the same time, these models are becoming spectacularly good at a large variety of tasks—3D design, coding, science, data analysis, biology, legal and medical analysis, research, etc. For these reasons, we actually think that the outlook for both Anthropic and OpenAI remains very attractive.

While many have been screaming “bubble” now for three years, even despite the massive cloud capex increases of the last years, each dollar of this capex so far has been justified and is actually making an attractive IRR based on our calculations. We’ll go through the math next.

First, the two frontier labs are currently doing $124 billion in ARR. If we gross OpenAI’s numbers—OpenAI strips out the revenue cut for cloud partners, who resell their models, while Anthropic includes these—we already get to $142 billion in ARR that end users are paying for these models.

At the same time, both OpenAI and Anthropic continue to show healthy growth while the capabilities of their models continue to improve at a brisk pace. LLMs are seeing increasing adoption in more and more end applications. Therefore, we estimate close to $187 billion in ARR that end users will be spending on these models by year end (total gross ARR line at the bottom of the model):

Second, AI returns are showing up as accelerating revenue growth inside the internet giants’ existing businesses. Macro pundits—who are generally very negative on AI—won’t be aware of this as they never actually listen to tech earnings calls. But a large variety of online apps have been discussing tremendous gains from better ad targeting, better ad placement, and better creatives thanks to AI. This includes also smaller ad networks such as Applovin and Unity.

However, the biggest gains were made by the big internet giants. For example, in Q2, Meta grew advertising revenue 27% year over year (ARR $237 billion) while Google grew search revenue 17% (ARR $253 billion) and Youtube ads at a 13% rate (ARR $44 billion). Both of these have regularly mentioned AI as the key contributor to their accelerating revenue gains over the past few years. For example, this is Zuck on the Q2 Meta call:

“First, we are now at a point where our investments in AI are accelerating every major part of our core business. They’re improving the experience for people using our apps, driving better performance for advertisers and helping our teams build new experiences and ship faster.”

And on the Q4 call of last year:

“Our business also performed very well, thanks to record-breaking holiday demand and AI-driven performance gains. We are now seeing a major AI acceleration. I expect 2026 to be a year where this wave accelerates even further on several fronts. We’re starting to see agents really work. This will unlock the ability to build completely new products and transform how we work.”

Same on Google’s call, this is Sundar:

“Our AI investments are redefining what’s possible across every part of our business. Our momentum starts with Search, where people are adopting one seamless search experience across AI Overviews and AI Mode. We saw 17% revenue growth in Search and Other and YouTube ads grew 13%.”

The numbers actually prove these statements. For example, Meta revenues were almost flat from Q1 of 2021 to Q1 of 2023. As from 2023—the year when transformer-based AI models really took off—Meta has steeply accelerated revenue generation. Basically, the company went from $114 billion ARR to $243 billion in ARR currently. That’s a $129 billion gain in ARR since 2023.

Assuming Meta’s benefits from new AI systems only kicked in as from 2024, we calculate a normalized ex-AI ARR for Meta based on normalized, mature growth rates. We estimate overall $178 billion in normalized ARR for Meta without the recent breakthroughs in AI. This suggests that AI added $65 billion in ARR for Meta over the past 2.5 years (the gap between the red and blue lines).

If we do a similar exercise for Google, we can see a similar picture. Quarterly ads revenues (Search + YouTube) grew strongly as from 2023—from a $188 billion ARR in Q1 of 2023 to $297 billion in ARR currently, a $109 billion gain.

Assuming AI benefits only really started kicking in as from 2024, we estimate $268 billion in normalized ARR for Google without the recent benefits from AI. This suggests that AI added $30 billion in ARR for Google over the past 2.5 years.

Overall, our best estimate is that AI has boosted Google’s current ARR with $30 billion and Meta’s with $65 billion. So, a boost of $95 billion in ARR in total for the two internet giants alone. This the part that most market commentators miss when they discuss the revenues and IRR that AI is actually generating.

Third, also clouds are bringing in substantial AI revenue. In the below table, we stripped out compute sales to frontier AI labs (Anthropic + OpenAI) in order to avoid double counting. In total, we get to $66 billion in ARR by year end:

Next, we have AI native apps and the tokens that SaaS is selling—again, we stripped out revenues paid to frontier labs. These add another $33 billion in ARR in total:

Overall, based on our math above, by the end of this year, we should already have $187 billion in ARR that frontier models are generating (split between the labs and clouds who resell their models). Then, ad-based ARRs for the two internet giants, Meta and Google, have been boosted by $95 billion over the past 2.5 years thanks to AI. We estimate this will increase to $108 billion by year end. Third, the various clouds—excluding frontier lab revenue—should be about $66 billion in AI ARR by the end of this year. Finally, also AI native apps and SaaS are generating $33 billion in AI revenue.

If we sum all of these up, we get to about $394 billion in ARR. There are a few smaller line items which we didn’t analyze above—such as other frontier AI (Grok, Gemini, etc.) and how AI is benefitting other online ads businesses (such as Amazon, Applovin, etc.). If we also include these, we get to $414 billion in AI ARR by the end of this year.

Taking reasonable assumptions, such as a 6 year life for the equipment, this should give an overall 12% IRR on the $931 billion in AI capex made over the past 3.5 years.

Contrary to many reports, we find that AI in aggregate is actually generating substantial revenues and is already generating a healthy IRR for the overall ecosystem. It doesn’t tell you however how this IRR is split. The big tech names such as Meta and Google will be capturing very attractive IRRs, whereas smaller cloud players which are highly leveraged—such as CoreWeave—might struggle to ever make an attractive IRR. As the market has figured out now, rising interest rates are a big negative for highly leveraged companies.

Nevertheless, the above math is a good barometer that both AI revenues and IRRs are already healthy—although some parts of it are hidden in the massive revenue growth which the big internet giants are seeing. This explains why the world remains heavily short on AI compute. For example, both internet giants and frontier labs are now paying neoclouds such as SpaceX and Nebius $40-60 billion per GW in order to get access to the compute they have available. A few years ago, neoclouds were typically signing $10 billion per GW contracts.

In the meantime, models continue to grow more powerful and use cases keep expanding—from coding to 3D design, mathematics, research, medical diagnosis, legal analysis, etc. This is Jensen on the BG2 podcast this week:

“The original GPT probably cost, pick a number, maybe $20 million to do if you wanted to train GPT-1. Now it will probably cost you $100 and we did that in 10 years. Meanwhile, our ambition got so large that the models that we are training now are 3 trillion, 10 trillion, 30 trillion parameters; we are talking about gigantic models now. Our ambition for the capabilities of the technology has gone up.

Our use of it has gone up probably 1 million times. I do not think this trend is going to stop. There is a fairly insatiable demand for intelligence. The smarter you are, the more intelligent you are, the more ideas of applying intelligence you have. The more I have learned in life, the more I realize there is more to do, because every time you learn something, it opens a new door. Every time you open a new door, you see more opportunity, you see more challenges, you see more problems.”

Finally, there is a massive opportunity to boost white collar productivity which we haven’t discussed yet. The global white-collar wage bill is roughly $35 trillion a year. We wouldn’t be surprised if in the coming decade or so, enterprises will start spending 10 to 20 percent of this budget on AI tokens. This would mean that the enterprise tokens TAM in the coming decade already could reach $3.5 to $7 trillion. To meet this demand, $2.5 trillion in annual AI capex would be required.

While many have been screaming ‘bubble’ for three years now—they must be getting really tired—based on the math above, we estimate that the revenue picture in AI already looks very healthy. At the same time, the outlook for further token demand growth is extremely strong. The notion that AI capex is going to collapse at some stage and that ‘this is the biggest misallocation of capital in history’ is in our view pretty foolish.

Many of the voices who have these views were also the same ones claiming back in 2016 already that tech was in a bubble based on a ‘very elevated Shiller PE ratio’. So, this is nothing new for us. Once again, we take the opposite view here and expect the trendline in AI capex to remain on an upwards trajectory over the coming 3-5 years.

We estimate that the $931 billion in AI capex over the last 3.5 years is already generating an attractive IRR for the overall AI ecosystem. Even if smaller cloud players such as CoreWeave, or other neoclouds, blow up due to excessive leverage, this available capacity will quickly get taken up by hyperscalers and frontier labs. The reason is simple—both of these remain heavily short on compute and still have tremendous token growth ahead of them.

Read more

Proteins are the key machinery to allow the human body to function. They make cells work (e.g. enzymes), they are the building blocks for the human body (e.g. collagen), they transport things around the body (e.g. hemoglobin carrying oxygen), they send signals (hormones), and they also defend the body against microbes (antibodies).

The below illustration from Duke shows how proteins are created. First, inside the cell nucleus, genes are read from DNA and transcribed into mRNA. This mRNA leaves the nucleus and goes out into the cytoplasm, the jelly-like fluid that fills the inside of a living cell, where a ribosome reads the mRNA code and then builds proteins.

So, the ribosome collects amino acids and stitches them together in the right order to build the protein. Once complete, the amino acid chain releases from the ribosome and folds into a functioning protein. This is exactly what Google’s AlphaFold AI model is all about—give it a sequence of amino acids and the model will predict the 3D structure of the protein. Before AlphaFold, it would take years of lab work to estimate the structure based on the amino chain sequence. The image shows a 3D rendering of a protein as predicted by AlphaFold and colored according to confidence:

Peptides are basically very small proteins, built out of much shorter chains of amino acids than the one shown above. A common rule of thumb is that anything under about 50 amino acids is a peptide, while longer chains are called proteins. Another key distinction is that while proteins fold into stable, complex 3D structures that determine their function; short peptides often have little fixed structure, although many still do important work.

Peptides have garnered a lot of interest both from the biotech space as well as on social media. Influencers on TikTok and Youtube regularly discuss the many benefits of particular peptides—from weight loss to faster healing, anti-aging, better sleep, more rapid muscle growth and even bigger erections. Clearly, these are very powerful molecules. However, note that many of these generic, consumer peptides are sold for ‘research purposes’, meaning that they’re not actually approved for human use. So, they can contain microbial contamination which could be life threatening. Secretary of Health Bobby Kennedy is now looking into approving some peptides for human use, meaning that you’ll potentially be able to get them at compounding pharmacies in the US going forward.

Peptides have also become big business for the pharma and biotech industry. OmniAb’s capital market day discussed how therapeutic peptides are now a $117 billion market which is projected to reach $333 billion by the mid-2030s, compounding at a 11% CAGR:

While the current engine of the therapeutic peptide market are GLP-1s, the peptides that make you lose weight, OmniAb shows that peptides can be used in a wide range of therapeutic markets. This includes big markets such as oncology (cancer) and cardiovascular, as well as high growth markets such as neurological and also rare diseases.

OmniAb provides a research platform to help pharma and biotech companies identify the right molecules to put through expensive trials. In return, OmniAb collects licensing and services revenues, as well as lucrative royalties on the sales of the drugs. Historically, OmniAb has been focused on antibodies but given the large potential in peptides, the company has now also developed a research platform to play in this new market.

This is OmniAb’s VP of Exploratory Research discussing the peptides research platform:

“I’m thrilled to tell you about OmniTides today. So we’re introducing this new platform, and it extends our reach into a high-value market that is adjacent to the antibody market that we have historically tailored our platforms towards. So as you’ve heard in the context of therapeutic antibodies, the immunization of animals is really a well-established and clinically proven method, and it has yielded most of the approved antibodies on the market today. We have a suite of transgenic animals that addresses that market.

What we’re doing is we’re repurposing the immune system. You could consider it nature’s discovery engine to discover peptides. And this really opens up new possibilities for us. Peptides have really been reinvigorated in recent years with the explosive use of GLP-1s, and these kind of medicines can transform diseases in different ways. So we really see peptides as being like a really exciting opportunity, and this is because of their small size. They’re able to penetrate tumors very effectively, and they have tunable clearance. For imaging, that clearance time frame can match the radioactive decay half-lives of radionuclides. This really expedites the discovery to lead times.

I want to share with you the OmniTides workflow. It’s really quite simple because what we do is we start with the OmniUltra chicken. So this is immunization. And what happens then is that it will produce these antibodies that have this protruding stalk-knob feature. You can think of this as an antibody that’s displaying a peptide that’s integrated into the antibody structure. We screen the output of an immunization campaign using xPloration.

So, having identified potential binders at the B cell secretion stage, we can then reconfirm these hits by recombinantly making them or we can also just make the peptide by chemical synthesis. This is pretty amazing that this little peptide that was discovered within the context of an antibody can actually function as an autonomous unit and it contains all the optimization that you would expect from antibodies.”

So, OmniAb uses genetically engineered chickens that carry human antibody genes. When these animals are exposed to a disease target, they produce fully human antibodies that can be developed into drugs. The animals’ immune systems generate a huge variety of antibodies and OmniAb’s xPloration tool selects the antibodies that bind best to the target. Then, the suitable peptides are cut out of these antibodies and chemically synthesized by manufacturing partners.

These types of research platforms can be tremendously powerful to find drug candidates. Regeneron is a good example here. Normally, if you immunize a regular mouse with a target protein, you get mouse antibodies. Human patients’ immune systems recognize those as foreign and so they have to be humanized afterwards, which is slow and can weaken the antibodies. Regeneron’s solution was to swap out a mouse’s own antibody genes for human ones, so directly inserting them into the mouse genome.

Earlier humanized mice had failed at this approach as fully human antibodies don’t communicate well with mouse immune cells. That led to weakened immune systems for humanized mice and so these mice weren’t effective at producing antibodies. Regeneron’s solution was to replace only specific parts of the mouse antibody genome with a human version. This is the part of the antibody that actually binds the target, while keeping the stem of the antibody—the part that interacts with the mouse’s own immune machinery—intact.

This made Regeneron’s mice both healthy and able to produce highly effective antibodies that also work inside the human body. A platform that allowed the company to launch one successful drug after the other. Investors were richly rewarded in this name—shares rose from $15 in 2008 to almost $1,200 in 2024.

OmniAb’s view is that chickens have an advantage over mice. Chickens are evolutionarily more distant from humans, making them respond to human proteins that mouse-based systems treat as self. Basically, chickens can produce antibodies for targets where mice fail.

Samsung Biologics is one of the largest drug manufacturers in the world. In biotech, these drug manufacturers are called the CDMOs (contract development and manufacturing organizations). CDMOs are similar to the foundries in the semiconductor industry. So, Nvidia will tell TSMC which chips to manufacture, while in biotech, a pharma or biotech name can outsource manufacturing to a CDMO. These CDMOs can produce a large variety of biological molecules as Lonza shows—biologics, cell & gene therapy, mRNA, etc.:

However, peptide manufacturing is a skill that few of the CDMOs possess. So, Samsung Biologics decided to step into this market by acquiring Polypeptide, one of the best peptide manufacturers in the world. This is Polypeptide’s CEO on what will likely be the last earnings call of the company, describing the attractive outlook:

“We posted in the first half of 2026 a strong growth in revenues and profitability. Our revenues increased 41.6% versus the first half of 2025 across both commercial and development. And this growth was driven by metabolics, which actually grew 72.6%. Our EBITDA margin improved from 2.7% in the first half of 2025 to 20.7% in the first half of 2026, and this reflects the benefit from the strong revenue growth of the company.

Our capacity expansion across our global manufacturing network is on track. Braine is now already operating at target capacity, and our capacity expansions across Malmo, Strasbourg and Ambernath are on track. We have further expanded our rich pipeline. The number of Phase III projects increased from 30 at the end of 2025 to 37 at the end of the first half of 2026. Just to put this into context, globally, the number of synthetic peptide drugs in development in Phase III is about 70. That basically means that about half of all Phase III development activity is going through PolyPeptide.

Peptide is one of the most attractive markets for CDMOs. The market is expected to quadruple from 2023 to 2031. When you look at the number of peptide drugs in clinical development, if you include early-stage activity, you have about 1,600 programs. Of those ones, you already have 591 in clinical development. And the clinical development is across multiple therapeutic areas.

Metabolic and oncology account for 40% of our clinical development activity. We are seeing a next wave of innovation in metabolics, which is the main growth engine for peptides. We are seeing indications expansion, whether it is MASH, cardiovascular, sleep apnea, osteoarthritis. We are seeing extended dosing intervals like monthly injections. We are seeing higher efficacy, better side effects profile. We are seeing alternative delivery routes like orals. And basically, what this is doing is enhancing the value proposition of metabolic peptide offerings, which will ensure that the strong demand that we are observing today continue for the foreseeable future.

If you look at the peptide market, there is always a mix between in-house and outsourcing. Some customers have in-house, some customers have 2, some customers rely more on outsourcing. Now based on what we see today, if you look at the number of pharmaceutical and biotech players moving to large peptide drugs without any peptide manufacturing expertise, this is very large. And that’s why we see they are relying on peptide CDMO support.

With concerns around biosecurity, when we look at Western customers, we are only competing against other Western players. We do not compete against Chinese players. Now at the same token, when there is a tender for China, we do not participate and it is mainly Chinese CDMO players participating. The concerns around intellectual property continues to be a barrier for Chinese CDMOs to participate more actively in the Western side. And of course, it’s a benefit for companies like PolyPeptide.

In the case of consumer peptides, we are following this development with interest. It could be an opportunity, but I have to say the demand from branded pharmaceutical and biotech players is so large that we are very much focused on bringing all these new products to market and supporting their commercialization. And generics is more or less the same. We do not really engage in terms of generic production. We believe most of the production is being done by Chinese and Indian players. But again, going back to what I discussed in terms of the level of innovation and development in terms of metabolic peptide drugs, we think that there are many differentiated offerings coming to market and the branded side of the market will continue to grow very rapidly.

Now, let me just talk about the Samsung Biologics’ offer to acquire PolyPeptide. As we communicated on the 20th of July, PolyPeptide and Samsung Biologics announced they have entered into a transaction agreement under which Samsung Biologics agreed to make an all-cash public tender offer to acquire all publicly held shares of PolyPeptide for CHF 44.31 per share. This transaction aims to provide PolyPeptide with the resources, investment capacity and a strategic platform to pursue the next phase of its growth and innovation. The combination of PolyPeptide’s deep scientific expertise and rich pipeline with Samsung Biologics’ industrial scale, resources and capabilities will create a transformational player to drive the growth in the overall peptide market.

Samsung Biologics is one of the most successful CDMOs globally and their track record and reputation is very high. So customers see this as an opportunity for PolyPeptide to access resources and capabilities to scale up in a way that before was going to be very difficult. And on that front, this is very positive. Our anchor shareholder supported the process and agreed to tender all of its shares, accounting for 55.65% based on total shares outstanding. And the Board of Directors, acting through its independent and nonconflicted members unanimously recommended that PolyPeptide shareholders accept the offer, a recommendation supported by an independent fairness opinion. This offer is a 40% premium over the undisturbed price before the market rumors of a potential transaction became public and it is a 60% premium on the VWAP on the 60-day volume-weighted average stock price, again before the news regarding a potential transaction became public.”

Polypeptide sees the global therapeutic peptide market growing at an 18% CAGR:

And the company revised its guidance up to 25-30% top line growth for this year:

Polypeptide is clearly a good company and the peptide space in general should be an interesting market for investors in the coming decade. Innovation is high here, as we saw with OmniAb’s discovery platform, which means that a lot of peptide-based drug candidates should go into trials. And as we saw with Polypeptide, companies in the peptide space are generating high growth and this is drawing strong M&A interest. As many countries are dealing with aging populations, peptides should be an interesting thematic for investors.

Next, we’ll review a company that we invested in a little while ago and that we reviewed for premium subscribers at the time. Shares are up 50% since, but the good news is that the outlook remains highly attractive. We expect these shares to compound at a 20-25% IRR in the coming years and with the potential to double in the coming 3-4 years.

The Best Way To Play The High-Growth Peptide Market

Read more

Compute Shortages & The Cloud

The central theme in AI continues to be supply chain shortages everywhere we look—all the way from the fabs and their suppliers to the data center. UBS notes that TSMC is accelerating its A14 buildout:

“Based on recent industry feedback and our analysis, TSMC is further accelerating the build-out of its leading-edge capacity. Specifically, the new A14 node should start tool move-in first at its Fab 20 in Baoshan from H226, and we believe the expansion scheduled at Fab 25 in Central Taiwan Science Park is being pulled forward to support larger capacity scale in 2028 when the node enters mass production. We estimate TSMC may build 60kwpm A14 capacity by end of 2028, above N2’s 30-40kwpm installed capacity in same stage. Looking across TSMC’s competitors, Intel targets high-volume manufacturing of 14A in 2028 with PDK 1.0 release later this year. Samsung Foundry appears to be delaying mass production of SF1.4 to 2029 due to yield challenges. We raise TSMC’s capex in 2027/28E from US $80bn/95bn to US$90bn/105bn to factor in more meaningful expansion across N3/N2 and faster ramp of A14 underpinned by Cloud AI demand and steady migration of smartphone & PC.”

Shortages in AI compute allow anyone who actually has the compute to opportunistically raise prices and boost margins—just like the memory players have been doing when it comes to semis—and both neoclouds and SpaceX are now taking advantage of this. Deutsche’s overview shows how SpaceX has been raising prices on their compute deals—from $31M per MW on their first customer to $50M per MW and then $60M per MW for their next customers:

SpaceX is printing cash on these deals and neoclouds such as Nebius and CoreWeave are now running the same playbook. This led JP Morgan to upgrade CoreWeave:

“We are upgrading CoreWeave to Overweight from Neutral, led by the favorable pricing backdrop for compute given the strong demand backdrop as well as the incremental willingness of the company to lean into short-term contracts at premium pricing, which drives our revenue and margin forecasts higher in contrast to a lackluster share price performance dictated by investor concerns around capital intensity.

The stronger pricing backdrop on account of a robust demand environment, which has strengthened through this year, is evident in the following drivers: 1) July pricing changes of 25% across SKUs; 2) frequent price increases across SKUs from peers like Nebius (not covered); 3) pricing for short-term compute contracts from some peer companies tracking almost 3x the pricing range for CoreWeave on longer-term contracts; 4) CoreWeave’s leverage of the opportunity evident in its recent press release highlighting contracts signed in F3Q at ~$40M/ MW; and 5) CoreWeave management highlighting that the robust pricing is already supporting 5-10 ppts of incremental contribution margins on contracts, relative to prior contracts, dispelling any concerns that the higher price is purely a pass-through of higher costs.”

It’s clear that there is currently a massive shortage in AI compute and that anyone who can make compute available can sell it at massive premiums. Last year, when we looked at neoclouds, the consensus was still that this capacity would be sold at a rate of around $10M per MW, which already would give a nice IRR. Currently, that capacity is being sold at $40-50M per MW by CoreWeave and Nebius, and even at $60M per MW by SpaceX.

Despite this improved revenue outlook, CoreWeave’s share price has gone nowhere, unlike Nebius which already has been selling capacity opportunistically to the highest bidder:

Truist shows how Nebius has been closing compute contracts with shorter durations vs CoreWeave (below). This makes the company a much stronger play on GPU pricing moving up, similar to SpaceX.

Another reason for CoreWeave’s underperformance is that the company is heavily relying on debt to finance its AI buildout. Nebius on the other hand has been relying more heavily on customer prepayments as well as dilutive equity offerings. While debt is attractive as it’s non-dilutive, the problem is that it’s also more risky, especially as interest rates are spiking due to the US-Iran conflict:

In the case of Nebius, 55% of capex is financed by customer prepayments, strongly reducing the risk. However, the advantage of CoreWeave’s strategy is that their debt financing has allowed them to close a massive backlog:

Overall, in the current environment, SpaceX and Nebius have the much better strategies. SpaceX raised $100 billion in its recent IPO, and can raise much more cash in the markets in the years ahead, to finance their AI buildout and train Grok. Nebius is similarly relying much less on debt to finance its cloud capex. And the big attraction is that both names have been selling capacity as it comes online to the highest bidders.

While CoreWeave has a huge backlog, all of this backlog is still at the old contract rates of $10M per MW. If the company brings more capacity online in the meantime, this can be sold at premium rates. However, this won’t be easy, as the company already has a massive debt load while interest rates are spiking. So, we think that it won’t be easy for CoreWeave to massively take advantage of the currently huge spreads when it comes to short term GPU capacity vs long term contracts.

Truist notes that also Nebius will likely have to raise substantial debt going forward to finance its buildout—obviously not great when interest rates are spiking.

So, if you’re bullish on AI compute capacity remaining in a shortage and GPU pricing continuing to move higher—which is a credible bull case given the huge demand for tokens combined with the NIMBY political backlash against data centers—the best play in our view is SpaceX. The company will be at a $100 billion ARR in the coming months already, has $100 billion in cash, is excellent at building data center capacity, and at current GPU rates they can get the payback in one year or so.

SpaceX trades at a massive premium compared to the neoclouds. However, SpaceX also has the extremely valuable crown jewels with Starlink and the launch businesses, while at the same time the company is competing at the frontier in AI with Grok. Grok is currently being trained on SpaceX’s internal data with the aim of making it the world’s best engineer. So, comparing SpaceX to the neoclouds isn’t really apples-to-apples.

A lower risk play on the above theme are the major hyperscaler clouds. While neoclouds mainly offer GPUs to power AI workloads, the major clouds offer a broad ecosystem of web services to run an entire business on.

Take ChatGPT for example—a neocloud can provide GPUs to purely power AI workloads, however, a major cloud can provide the entire infrastructure to run OpenAI’s business and the ChatGPT app on, including databases, analytics, observability, tons of CPUs, data centers around the world close to end users, etc.

As moving data out of these major clouds is both costly and slow, major clouds are able to price their GPUs at premiums as Truist’s analysis shows:

The big cloud players also have strong operating cash-flow generation to finance their AI buildouts, massively reducing the risk-profile. So, for investors looking for quality companies to get exposure to the growth in AI, we continue to see the major clouds as solid long term investments.

Valuations for these names remain attractive and so we have large positions in Amazon and Microsoft, with a smaller position in Google. We’ve recently also added to the two best clouds in China—Alibaba and Tencent—as these are ridiculously cheap in our view while these are actually very good businesses.

Synopsys & AI Chip Design

Read more

Alibaba held its annual flagship cloud event in Hangzhou this week and the big takeaway for us was that scaling laws are alive and well. AI training times will continue to increase with recursive self-improvement and increasing model sizes. These are the highlights from Eddie Wu’s keynote:

“Consider the early days of electricity: It was initially used only for lighting. In 1882, when Thomas Edison switched on the Pearl Street Station, it powered barely four hundred lightbulbs across a few blocks. Interestingly enough, in the beginning, Edison bundled light bulbs with free electricity, much like how AI agent platforms give away free tokens today. But the inventions that fundamentally changed human life were born much later. Air conditioning appeared in 1902; washing machines and refrigerators entered households years after that. And the first digital computer didn’t arrive until 1946—over six decades after those 400 lamps. Of course, in the age of AI, this evolutionary clock is running exponentially faster.

In the age of Machine Intelligence, powerful infrastructure is equally non-negotiable, and rests on three cornerstones: AI models, AI chips, and the AI cloud. Parallel to the historic rollout of electricity, the demand for AI infrastructure in the era of Machine Intelligence is practically limitless. Alibaba remains committed to building this core infrastructure across AI models, AI chips and the AI cloud as a long-term strategic priority.

AI researchers have also mapped out a concrete path forward: Recursive Self-Improvement (RSI). By engaging with real-world tasks and feedback, models identify their own limitations. They autonomously design experiments, synthesize data, and evaluate outcomes, driving a continuous cycle of self-evolution. Currently, Alibaba’s Qwen team is exploring RSI and has made meaningful progress. The team is continuing to advance research on model architecture and data optimization, and plans to train a new model at the scale of 5 to 10 trillion-parameter, with the goal of completing more complex, longer-horizon tasks and advancing toward ASI.

The second cornerstone is the AI chip. If tokens are the electricity powering the AI era, then chips are the power plants. In the upcoming age of Machine Intelligence, demand for tokens will be virtually boundless, requiring chips to continuously enhance performance and scale supply. Today, T-Head (the semiconductor subsidiary of Alibaba) is building a comprehensive portfolio of data center chips: the “Zhenwu” series for GPU chips, the “Yitian” series for CPU chips, the “Panmai” series for smart NICs, and ICN interconnect chips. The portfolio fully covers the core chips required to build ultra-large-scale AI clusters.

We are introducing our next-generation AI chip, the Zhenwu V900. It is the most powerful AI chip in China today, delivering three times the performance of its predecessor, the Zhenwu M890. A single cluster built on V900 can support up to 500,000 cards to power frontier model training and inference. Backed by the proven maturity of T-head product lines and widespread adoption across our clients, we anticipate a significant growth in the annual AI chip shipment volumes.

However, we recognize that the industry’s mid-to-long-term demand far outpaces our supply capabilities. Global shortages across the AI data center supply chain are currently limiting the speed at which we can scale our compute infrastructure. Our target is that by 2032, the global data center capacity operated by Alibaba Cloud will surpass 20GW, fueling the industry’s exponentially rising demand for AI.

Let history be our guide. When the Industrial Revolution transferred physical toil to machines, humans were liberated from grueling labor. When machines shoulder more of what must be done, humans will have the time to pursue what they truly desire to do. Looking ahead, Machine Intelligence will unlock unprecedented space for human curiosity and creativity. Just as a farmer working in the fields three centuries ago could scarcely imagine that people would one day pay to perform physical exertion in a “gym,” we, too, find ourselves constrained in fully capturing the landscape of what lies ahead. This defines the ultimate purpose of our commitment to the era of Machine Intelligence: to offload onerous tasks to machines while preserving time, creativity, and the appreciation of life’s beauty for humanity. All of this is only just beginning.”

Over time, the cloud and AI will become increasingly important in Alibaba’s revenue mix. Currently, cloud is only 15% of revenues, but over time, when Alibaba has 20GW or more of capacity, the cloud and AI will become dominant in the revenue mix. Especially if they can also start monetizing their Qwen models. For example, in China, demand for non-Western models should be high.

A recent blog post by OpenAI’s chief scientist made a similar conclusion—we’re only at the start of digital intelligence and models will continue to get smarter with scaling and RSI:

“Based on internal results, I have a strong expectation that this speed of progress could be sustained into recursive self-improvement. If AI development continues along its current path, the systems we’ll see in the next few years are likely to represent further capability jumps of equal or larger magnitude, and to increasingly drive their own development. This is a time that calls for extreme caution. I am concerned no one is prepared for the consequences of a continued rapid rise in machine intelligence.

At a high level, progress in machine intelligence is driven by increasing computational power. We at OpenAI deeply internalized this around 2017, after seeing consistent returns to scaling across multiple research projects. As a result, we sought out access to much more compute than we had originally planned, and increasingly oriented our research around a small number of very scalable directions. We believed that was the only way for us to be at the frontier of AI research, and influence the impacts of AGI.

There are new algorithms that have been developed along the way, new feats of ingenuity from teams and individual researchers. I see them largely as discoveries along the path of scaling; the science of deep learning is still nascent, and meaningful algorithmic progress tends to correlate with access to compute. If you zoom out to a multiple-year horizon, AI is continuing to become more intelligent as it is scaled to larger computers.

We spend a lot of time trying to understand how capabilities generalize, and what to prioritize to advance the skills that are going to be most relevant in the next few years. For instance, we believe we could make the models better at specifically mathematics research with additional focus, but we do not prioritize this direction because of the urgency we feel about RSI and automated alignment research.”

A problem for Alibaba, however, is that China is heavily short on advanced compute as they don’t have access to EUV. Although also China is clearly innovating when it comes to advanced semis. For example, we came across this interesting analysis from Bernstein concluding that Huawei’s latest 7nm-like chip outperformed an Apple 3nm chip due to a superior chip architecture. As reported by the SCMP:

“Huawei Technologies’ latest smartphone processor, the Kirin 9050 Pro, unveiled earlier this month, has narrowed its gap with Apple’s chips to about three years, from roughly four years in the previous generation, according to equity research firm Bernstein. The chip, which Bernstein estimated was produced using technology equivalent to a 7-nanometre process node, outperformed Apple’s 3nm A17 Pro in Geekbench 6 multicore tests, a widely used benchmark that measures how effectively processors handle tasks across multiple central processing unit cores.

The research firm estimated that the Kirin 9050 Pro was manufactured using SMIC’s “N+2” or “N+3” process technology, equivalent to a 7nm-class node achieved through deep-ultraviolet lithography, rather than the more advanced extreme-ultraviolet tools unavailable in China due to US sanctions. At the centre of the progress is LogicFolding, a technique developed under Huawei’s Tau Scaling Law framework that stacks logic vertically and uses dense interconnections between layers. The approach reduces the distance signals travel within a chip instead of relying primarily on shrinking transistors. According to Bernstein, the technique increased transistor density to 238 million transistors per square millimetre from about 155 million – a gain of more than 50 per cent.”

Now that the Chinese government has ended the severe crackdown on big tech while recently also stopping the price wars in food delivery between the various apps, e-commerce should remain an attractive cash cow for Alibaba. And then with the boom in token demand, the strong growth of Alibaba Cloud (+45% yoy in the latest quarter) should make Alibaba a growth stock again.

Note that Alibaba Cloud is a major cloud platform in China, similar to Amazon Web Services in the US. And unlike its Western peers, with the exception of Google, Alibaba actually has strong LLM capabilities with its Qwen models consistently being highly ranked among the open-source models. Proprietary LLM capabilities is really an area where the big two Western clouds, Amazon and Microsoft, have been lacking.

Alibaba is a cheap stock on 14x forward EPS. For people looking to invest in China and in great businesses that can grow with AI, we suspect that these shares will do well in the coming 3-5 years.

Next, we’ll review the outlook and our findings for the following topics:

• Hybrid bonding & Besi’s sell-off

• Power semis

• TPUs in space & the best way to play

As we argued back in May-June, valuations got way too stretched in semis. A lot of fast money moved into the space, driven by the high momentum in semi share prices and with hopes of making quick gains. We were net sellers of semis in that period and locked in some nice profits—we basically halved or moved entirely out of the more expensive names, while keeping positions in names which still looked cheap.

At the same time, we purchased names such as Snowflake and Zscaler which were cheap at the time and which looked like obvious winners from AI to us. The consensus was that software would get disrupted due to vibe coding, but it’s not that simple, and as we predicted, growth actually strongly accelerated for our top picks in software such as Snowflake and Unity.

Hybrid Bonding & Besi’s Sell-Off

Read more

Elon gave a number of indications of the amount of capacity SpaceX will be deploying:

“We expect to end this year with over 2 gigawatts of compute. And probably our cumulative compute online by the end of next year will be several times higher. So it may, let’s say, be closer to 10 gigawatts of compute than 5 gigawatts of compute.”

And the company is looking to far exceed that number in terms of available power they’re planning to have online; which is the best indicator for how much capacity they’re planning to add subsequently:

“So we’re actually aiming to far exceed that gigawatt number in terms of power online, power cooling and electrical equipment. So our tentative target is to actually have 20 gigawatts at the power and cooling level online by the end of next year. Now I don’t think we’re going to achieve 20 gigawatts, but we want to have a series of projects that cumulatively come to 20 gigawatts by the end of next year. Some of them won’t pan out exactly on time, but I would expect that we still probably have at the power plant level, something close to 15 gigawatts. Assuming that maybe 1/4 of the projects take longer than expected. So our goal is to have far more power, cooling and electrical equipment than we have GPUs. That’s a logical thing to do given the relative expense of GPUs versus balance of system.”

So, SpaceX will probably install around 6GW of AI compute in ‘27 and then another 8GW or so in ‘28. This gets us to 16GW of cumulative AI capacity by the end of ‘28.

We also know that SpaceX has Nvidia as its exclusive provider and that Nvidia makes $25 billion per GW on Blackwell and $40 billion per GW on Vera Rubin. So, combining all numbers above, we can work out that SpaceX will contribute around 35% to Nvidia revenues in the coming years:

Some have speculated that Nvidia has given discounts in order to obtain this exclusive partnership, however, as Nvidia guided for 72-73% gross margins in calendar ‘27, we know that they aren’t doing much discounting. So, we think that Jensen simply ensures that SpaceX gets the hardware whenever they need it. And as soon as Jensen has the latest product available—Vera Rubin currently—we can also be sure that SpaceX is a preferential customer who gets first access.

Using the same methodology as above, we let OpenAI Astra estimate Nvidia’s full revenue split for calendar 2027 based on what each customer has been announcing in terms of capacity additions on its latest calls. Astra’s numbers for SpaceX are fairly close to ours:

Thus, Nvidia’s outlook in the coming 3-5 years is becoming reliant on the success of SpaceX. However, long term, it’s well possible that SpaceX will be looking to design its own data center silicon as well. Musk’s other business, Tesla, is already designing its own inference chip for self-driving and Optimus, and has in the past also designed its own Dojo data center chip.

SpaceX and Tesla are even looking to build their own ‘Terafab’ to build their own memory and silicon. The initial R&D fab will be a unique building in the world, where the full loop of all manufacturing processes for a modern chip package can happen under one roof. The goal is very fast product iteration. This is Elon discussing how SpaceX-Tesla is looking to get a competitive advantage when it comes to silicon design:

“We’ve placed equipment orders for our development fab in Austin. And that development fab is pretty cool because it’s intended to have lithography mask production, logic, memory, packaging and chip testing all under one roof. So you can have a very fast iterative cycle and try out new chip designs very quickly and see if they work. I don’t think such a building exists anywhere on Earth. So this is really going to be super helpful as we try some exciting, adventurous, high-risk, high-payoff bets on AI chips.”

Broadcom’s Sandbagging?

After the breakdown of Nvidia’s revenues per customer, we then tried a similar exercise for Broadcom. Also here it’s actually fairly easy to get the customer mix as Hock Tan told us what they are looking to ship in ‘28:

“Moving on to Anthropic, starting with the 1 gigawatt of Ironwood we are deploying in 2026. We expect Anthropic to deploy another 5 gigawatts of TPU version 8i in 2027. And in 2028, we have clear line of sight to deliver an incremental 10 gigawatts, even as we expect Google to grow for us. Anthropic is on track to become our largest XPU customer in 2027 and sustain that in 2028.

For OpenAI, Jalapeno is on track for the planned deployment of 1.3 gigawatts in 2027. Together with OpenAI, we are deep in development of the next-generation XPU beyond Jalapeno, which is approaching tape-out. In 2028, we have line of sight for OpenAI to deploy over 5 gigawatts of Jalapeno and its successor generation of XPU, which would make OpenAI our second largest XPU customer. In addition, we are in development with OpenAI on their third-generation XPU. As OpenAI announced last week, Jalapeno outperforms the Grace Blackwell Ultra in performance per watt, latency, throughput and power and it’s actually comparable to Vera Rubin GPUs in running OpenAI workloads. The lesson here is when you co-develop a chip that is optimized for your particular LLM workloads, you will outperform any GPU. Like TPUs, Jalapeno demonstrates that it can also run other frontier models and you can do all this at half the cost of a GPU.

Our partnership with Meta to deliver multiple generations of MTIA XPUs remains on track. Between now and the end of 2027, we will be delivering 3 generations of MTIA accelerators to Meta. Across these 3 generations, we have line of sight to deploy 3 gigawatts through ‘28.

Our content in AI, as you know, goes beyond XPUs. We’re the leader in AI networking, and we continue to extend our lead. In sum, we continue to invest and invest heavily to provide the broadest and most leading-edge AI portfolio. In fact, our AI networking revenue is expected to grow just as fast as XPUs over the next few years.”

If we add all those numbers up, we can see that Broadcom is planning to ship 17GW in ‘28 (Anthropic: 10GW, OpenAI: 5GW and Meta: 2GW). Broadcom recently guided to be making “$20 billion to $30 billion” in revenue per GW. So, if we take the midpoint here at $25 billion per GW, we arrive at $425 billion in potential AI revenue in ‘28. The company is only guiding for $230 billion in AI semiconductor revenue for that year. Based on the bottom-up GW shipments they’ve disclosed, there is a lot of upside to overall guided revenues.

And the above bottom-up numbers don’t include the shipments to Google—if we add in Google as well, we get to $475 billion in potential revenue:

Whereas the dominant debate topic over the last year for Broadcom has been the company’s market share in upcoming TPU generations with Google, the bigger question is how much AI silicon the company will be shipping to Anthropic and OpenAI in the coming years. The overall risk is that there will be delays in shipments, however, based on the above math, it looks to us that Broadcom is building in a lot of conservatism in their guidance and with substantial opportunity to surprise on the upside.

The above calculations make it very clear that Nvidia is becoming heavily exposed to the neocloud market, and especially SpaceX, whereas Broadcom has very high exposure to Anthropic and OpenAI. Thus, who will win the race in frontier AI is becoming a crucial question for semi investors. If Grok 5 will blow everyone out of the water or if open-source models take token share, this is good news for Nvidia. However, if Anthropic and OpenAI continue to dominate the market in terms of revenue share, this is obviously a massive positive for Broadcom.

Out of the two, our top pick would be Broadcom here. In our view, OpenAI and Anthropic are by far the best positioned companies when it comes to monetizing frontier AI, and Broadcom is the clear play on these two building out AI infrastructure. We continue to hold Nvidia as well but we’ll reduce our holding here, and then increase the position in Broadcom.

Next, we’ll dive into the following topics:

• Credo vs Marvell in Scale Up Optics

• A detailed update on SpaceX

• SpaceX vs ASTS

Credo vs Marvell in Scale Up Optics

Read more

Nebius made an interesting bull case for open source models this week. Basically, enterprises don’t trust the frontier labs with their most sensitive data and so Nebius is seeing demand from enterprises to fully control their own models and infrastructure:

“Customers like big corporations, they are concerned that when they use AI, they need to feed their data back to other companies. They kind of give up all of their secrets to somebody else, which then they incorporate in the universal models and then can be used by competitors to these enterprises. So they would rather see a model which they can control and in an infrastructure they know, and that they can look through to the ground. And Palantir actually have chosen us because we provide this full stack infrastructure which they can have full control of. On top of it, they put their software.

Our token factory provides all kinds of token models—open models, Chinese models, Nemotron and others. Palantir has instruments for the enterprise—which allows them to take their data, take an open model, feed the data to the model, the model generates outcomes, and then they use it as their new data. They feed it back to the model. And after several cycles of training with your own data in your specific domain, it becomes higher quality than the universal model. And that’s what happened with Shopify.

They started with open model Qwen and after several cycles, they reached the level of quality in their specific domain, higher than GPT-5.6. So this cycle—open model and your own data—helps to generate intelligence, which is narrow domain smarter than the Super Duper universal models. That’s the thesis of Palantir. We fully support it, NVIDIA supports it. Probably this is the way to go, or one of the ways to go. Of course, universal models will not disappear. They will have their place. But this is the thing which Palantir wants to do. That’s our partnership.”

This actually makes a lot of sense. For example, if you’re ASML, are you going to run all your proprietary know-how on EUV technology through Claude or ChatGPT? To be honest, we wouldn’t recommend doing that. It makes a lot more sense in this instance to take an open-source model such as Nvidia Nemotron, deploy it in an infrastructure you can fully control—ideally, your own data center even in the case of ASML—and then you can be fully confident that the AI which is assisting you isn’t going to leak any of your biggest secrets.

For example, we were coding over the summer an app with Opus and Opus noted that a setup we were proposing for the workflow was a very clever idea. This month, with the latest Fable 5.1 update, we noticed that Fable proposed a similar setup on its own for another app we were coding. Now, Fable is smarter than Opus, however, it’s also possible that there is a backdoor when you use these models. Claude might have a system prompt along the lines of—“if the user has a very clever idea, or a solution or specific knowledge you haven’t encountered before, send this information over API to this particular database”.

Alternatively, Anthropic can also have Claude read all customer interactions to gather new capabilities. Anthropic obviously does this anyways, this is what they wrote on X a few days ago:

“We’re publishing our most detailed threat intelligence report to date. It covers how people tried to misuse Claude—for cyberattacks, influence operations, surveillance, biology, and building weapons—and how we found and stopped them…”

They’re obviously analyzing everything that Claude is doing and understandably so. Business-wise, it makes sense to use Claude to gather the best ideas and new knowledge from customer interactions so that Anthropic can train the best possible universal model. Being at the cutting-edge frontier in AI capabilities is what generates revenue. It’s the difference currently between a $65 billion ARR and a $1-2 billion ARR. Naturally, this is pure speculation from our side, and we have no proof that this is actually happening. But business-wise, it just makes sense and it would be fairly easy to do this behind the scenes for both Anthropic and OpenAI.

Broadcom’s CEO, Hock Tan, pointed out at the Goldman conference that the current economics of open-source are still terrible, and likely not sustainable:

“The total amount of compute tokens consumed and generated, the cost of generating those tokens to be consumed globally, and this data is obviously for inference, it’s around $200 billion a year now, give or take, rough numbers. That’s the cost of generating those tokens, producing those tokens in infrastructure. Ask yourselves, what’s the revenue you could attribute being created and earned by those model guys. I give you that, around $150 billion. Interesting, isn’t it?

This is generative AI today. I know it’s early stage, things are still growing, but we’re generating $200 billion of token costs to generate $150 billion. But, look below the surface, you can split it up. On those tokens generated, half of it is generated through frontier models, closed systems. The other half generated through players offering open-source, open-weight models. So that’s about it. But you look at the revenue and 75% at least is coming from frontier models. That’s $120 billion, and they spend roughly $100 billion. It’s not so bad and it’s still growing. Let’s look at open-weight models. Spending the other $100 billion, they generate $30 billion revenue. Do you think that’s a sustainable model? We don’t know, but it proves one thing. The value goes to where intelligence continue to improve.”

When we look at Z.AI’s financials—the creator of the GLM open source model—we can see that the company burned $323 million in cash over the last twelve months to generate $219 million in revenues. However, we’re still early, and the economics of these models will improve over time as these companies are looking to introduce commercial licenses for enterprises to adopt these models.

Broadcom is also talking their own book here. They’re really the play on closed-source winning as basically all leading frontier labs are increasingly designing their own silicon with Broadcom as their key partner. At the other end of the spectrum, Nvidia is really the play on fragmentation in AI. If enterprises fine-tune and retrain a variety of open-source models, all these various models will need a uniform compute platform and hardware to still get scale advantages in infrastructure.

We see both these scenarios playing out in the coming five years, with the most likely scenario being that token volumes will be split between frontier and open-source models. For enterprises, it makes sense to fine-tune open-source models where they have to handle sensitive data. However, when you’re writing a codebase or doing engineering work where none of what you’re doing is particularly secret know-how (e.g. building a CRM SaaS), it’s clear that both ChatGPT Astra and Claude Fable are state-of-the-art solutions. Therefore, we continue to see both Nvidia and Broadcom as two attractive plays in data center semis.

While the market is currently worried about Broadcom’s market share with Google in upcoming TPU generations, Macquarie points out that ASIC demand from OpenAI and Anthropic will surpass that of Google:

“Our earlier caution rested on Google insourcing its TPU and diversifying to MediaTek (see details here). We believe that risk has played out and is in the price as the shares corrected ~24% from the 2026 high, and Google’s direct investment in MediaTek confirms the shift.

Mgt said Anthropic is on track to become Broadcom’s largest XPU customer from FY27 and guided AI semi rev to double in FY27 and FY28. We forecast Anthropic to purchase >US $40bn from Broadcom in FY28, more than offsetting the Google loss. With Anthropic’s IPO pending, we think Broadcom is the cleanest listed way to own the build. We acknowledge what is secured and treat the FY28 outlook beyond that as timing optionality. Component tightness shifts deployments right, not away, so catch-up toward the guide is upside to our numbers.

Our earlier concern was reliance on a single client (i.e. Google). Growth now rests more evenly on six XPU customers. Mgt sees OpenAI as its second-largest XPU customer by FY28 (>5GW of Jalapeno and its successor) and Meta shipping three MTIA generations (3GW through FY28), while AI networking grows as fast as XPUs.”

We see Broadcom as an interesting play on the S-curve in frontier AI in the coming 3-5 years, also as these labs will increasingly shift volumes to their own ASICs:

The bull case for Nvidia is that open-source models will continue to have a substantial share of overall token demand, which we think they will, making Nvidia an interesting play on overall annual token growth at 18x EPS. In our view, the bull case for Nvidia is playing out as the company recently upped their guidance again and is now looking at around 70% revenue growth for calendar 2027.

Next, we will discuss our finds on Snowflake and robotaxis.

Read more

“Initially when people started working on models for language, the dominant design was LSTMs. Some people remember what those are. They were kind of okay; they were a lot better than what came before that, but they didn’t really scale as well. And then the big thing with transformers was not that transformers were somehow particularly mathematically elegant or anything like that. It’s just that they scaled better. So they were easier to train on very large amounts of data with lots of parameters. The technology proceeds in phases. First you figure out what you can scale, and then you pour on a lot industrial-scale effort and adding lots of data, adding to model size. And that’s when the magic happens.

What I would say about robotics is that it’s not in the GPT-4 to GPT-5 stage, where it’s an industrial-scale effort to make the model bigger and get more capability out of it. It’s in that stage where we’re establishing the fundamental technologies. And because of that, what one should expect to see right now is not necessarily that each month the model gets bigger and more powerful by some predictable kind of scaling curve. It’s that the scaling properties themselves are evolving as we develop the right technologies.

So, where we’re at now is that we’re actually getting all those puzzle pieces in place. And I think it’s actually very close. I think a lot of the puzzle pieces are falling in place, but it’s not yet at that predictable scaling stage. Certainly we’ve seen a lot more happening as we progress, but they were kind of magical because in robotics, basically prior to 2024, the stuff never happened. So the little things that happen, and this was maybe at this point about two years back, we would see things like, okay, we train our policy for folding laundry and it takes individual shirts out of the hamper and tries to fold them. And then one very vivid memory I have in late 2024 is we were watching one of the evals, and it takes out two shirts at the same time. And I’m watching this, I’m like, okay, it’s done for. There’s no way you can possibly do this. And then it puts the two shirts on the table, disentangles them, puts one of them back, and then starts folding the other one. It’s like, wow, okay. In retrospect, you can do some detective work and figure out where it got that from, some piece of training data. It’s exhibiting the common sense you expect people to have.

What would have the biggest impact on the advancement of robotics. Certainly, availability of reliable, low-cost hardware is a big deal. Right now, hardware used for robotics research, a lot of that does come from China. And it’s good, it’s relatively inexpensive, it’s of high quality, and meets the standards that people generally need. But it would be awfully nice to be able to source all that domestically as well. Having an effective positive feedback loop where more deployed robots translates to more model capability, that’s kind of the key. The hard thing in robotics was always generalization. What is very easy to miss in the hype cycle and the demos that people put out, is that when somebody shows a demonstration of their system, the demonstration alone usually doesn’t make it clear what level of generalization is being shown.

The highly acrobatic robot demos, for example, are really exciting to look at. But typically, it’s a show that is obviously rehearsed. It’s not the same as doing a task every time reliably in any home. So you might see the robot doing something fairly mundane and unimpressive, but what’s exciting about it is that it’s doing it with an object that it’s never seen before in an environment that has never been tested before. And that is actually harder than doing an acrobatic backflip that it’s practiced millions of times. One of the steps on the roadmap is to have a robotic system that gets better with autonomous experience in a setting that it wasn’t originally trained for. I get my model, I get my adaptation algorithm, and I put it in a new setting. Maybe it’s a home, maybe it’s a factory, something where it’s doing something real, and it does okay. But then over time, it gets better and better. That would be a major milestone.

Live demos with really long time lapses, I think it’s a great idea. We had our robot assembling boxes at Dandelion Chocolate Factory. So we ran it for several days. But we had this coffee task, which was the robot using an espresso machine to make espresso. We ran it for 13 hours making espresso drinks. It screwed up a few times. It’ll spill all the coffee grounds, and then it needs to go and get a cloth and wipe it down. The way that experiment was done is that there is a high-level prompting. So the prompt is updated roughly every five minutes or so in between semantically coherent tasks. So you tell it: make espresso, clean up the machine, etc. So those steps, the actual clean up the machine, was commanded by a person. In principle, we could automate that. In fact, one of the things we’re spending a lot of effort now on is improving our high-level policy that does those commands. But for that experiment, every five minutes, somebody basically updates what it’s being asked to do, the way we intended. It was like the commands would be like, if you go to an actual coffee shop, you say, “Oh, I want a latte, I want an espresso.” That was supposed to be the prompt. Except then you also have to tell it, “I want you to clean it up before you do the next one.”

One thing that I’ve learned over the last few years is that addressing robotics effectively requires using very broad prior knowledge. For example, I was very skeptical initially about the utility of language. And I think scientifically this is defensible, which is that animals can do some pretty impressive things, but they can’t speak, at least not very eloquently. So maybe our robots should also be able to do stuff, and they don’t necessarily need to understand language. But what’s important is prior knowledge that you can put it in as a scaffold on your learning process. And you can pull in that knowledge in all sorts of ways.”

If we look at the progress in AI—or machine learning, as it used to be called a decade ago—we’re probably in a similar phase now in robotics as we were in digital neural nets one decade ago, around 2016 or so. This was a time when AI had a good understanding of what is in an image, what’s the meaning of a text, and reinforcement learning had just been introduced to give AI superhuman performance. Currently, while robots start being able to do basic tasks, we’re not in a phase yet where they have as good an understanding of the surrounding environment as humans do.

A few months ago, on the Invest with the Best podcast, Levine gave more background on his company and the robotics field:

“Part of the thesis of this company is that we believe doing it at the full level of generality might actually in the long run be easier than trying to special-case very specific narrow application domains. Again, in much the same way that for language models, it turned out to be easier in some ways to solve natural language tasks in their full generality than to narrowly target machine translation or sentiment analysis. In the world of natural language, we saw that there were a lot of efforts to develop domain-specific solutions that tackled specific problems.

For example, somebody would spend a lot of time thinking about how English differs from French and then build a machine translation system. The reason that language models actually took over for all of those different application domains is because they can leverage much broader sources of data. When you can leverage weakly labeled data, like data that you mine from the web in the case of language models, you actually learn more about the world. You establish a foundation of world understanding, and on top of that foundation, it turns out to be much more effective to build out different applications.

To bring this into robotics, the calculus does not look quite the same because in robotics we do not have an internet-sized dataset that we can just draw on. But this notion of understanding the world is actually more important in robotics. If you have many different tasks and many different physical systems, then you can go from training individual dishwashing specialists or laundry-folding specialists to instead training a model that actually understands physical interaction. People can master new skills very rapidly because we understand physical interaction. We can intuitively grasp what is going to happen in a new unfamiliar situation, which lets us bootstrap things really quickly. If we can draw on data from many sources, many applications, and many robots, then we can have a model that has physical understanding, making it much easier to put new applications on top of that platform.

The point of generalization is that it does something relatively mundane that any human could do, but it does it in any situation. We released some demos where we showed our robot cleaning kitchens. If you watch an individual video out of context, it just looks like it is picking up plates, which anybody can do. But the point was that we put the robot into that home just for that demo, and it had never seen training data from that setting. You have to understand what is going on behind the scenes to appreciate why that is pushing the frontier.

In the future, we will have a robotic foundation model that can be adapted to all sorts of applications, running the gamut from bulldozers to humanoids to robotic arms. The fundamentals of how you interact with objects, how things move in the world, and how causality works are conserved across all of these different systems. Machine learning works best when there is a lot of data. If you naively approach a problem like washing dishes by collecting an enormous amount of dishwashing data, it is not cost-effective because you have to repeat that entire process for the next task. Training general-purpose models reduces the data needed per task.

What has changed most recently is the ability to handle unusual scenarios. In edge cases where you lack prior experience, you must rely on knowledge acquired from other sources and ground it in the new situation. Multimodal language models are adept at pulling in broad web knowledge. While they are not inherently grounded in physical situations, they provide a path to import common sense into robotic control. The first deep reinforcement learning systems in the early 2010s were a major milestone because deep RL gives us a way to surpass human-level performance. More recently, the advent of multimodal LLMs adapted for robotic control to supply common sense is a critical advance, and we will likely see several more major breakthroughs over the next few years.

Over the past few years, we started by developing Vision-Language-Action (VLA) models. A VLA model can be understood as an LLM adapted for robotic control. It is pre-trained on text, adapted with extensive web image data to understand visual scenes, and then fine-tuned on diverse robotic data. That serves as the foundation for bringing web-scale knowledge into robot control. On top of that, we explore two primary capabilities: handling unusual situations with common sense, and improving through reinforcement learning. To achieve common sense, we utilize chain-of-thought reasoning. When the robot encounters a scene, instead of immediately executing motor commands, it reasons through the task. If asked to clean a kitchen, it analyzes the visual input, generates an intermediate semantic thought like picking up a specific plate, and then acts. These intermediate inferences benefit from web-scale pre-training.

Reinforcement learning is applied when practicing tasks repeatedly to improve robustness, speed, and precision. In our espresso-making demonstration, the system practiced repeatedly to optimize throughput and reliability. We are continuing to build upon this dual foundation. Nobody knows exactly how much robot data is needed to achieve universal generalization, but we may not need to know in advance. The objective is to deploy systems capable of performing diverse tasks and establishing a data flywheel.

Humans intuitively assume that tasks difficult for us, like advanced calculus, are hard for machines, while tasks easy for us, like grasping a cup, should be simple. In reality, human brains have evolved massive specialized machinery for physical interaction and visual perception. Machine learning shifts this dynamic. Where data collection is straightforward, tasks become tractable even if they are physically intricate. While learning is now widely accepted, there is still debate over end-to-end learning versus modular systems incorporating hand-coded physics. The Bitter Lesson posits that leveraging general methods with data and compute outperforms hand-engineered domain knowledge over time. While incorporating explicit physical models offers short-term structure, end-to-end learning provides the scalability and self-improvement necessary for true open-world generality.

Former Everyday Robots researcher Benji Holson proposed a practical Robot Olympics consisting of everyday manipulation tasks that humans find trivial but robots struggle with—such as opening doors, scrubbing greasy pans, or using a bag to pick up waste. We used this list to test our general onboarding pipeline. Without developing specialized algorithms for individual tasks, our foundation model successfully solved nearly all of them. The only exceptions were turning a dress shirt inside out, which was physically constrained by gripper size, and peeling an orange purely with fingers, which required a small tool due to grip strength limits. This demonstrated the power of a general model to onboard diverse tasks rapidly. Tasks expected to be among the last to be automated by robots involve direct, nuanced physical interaction with humans such as changing a child’s diaper, eldercare, or helping an injured person out of bed. These tasks require precise force modulation, high safety margins, and deep social-physical common sense.”

While it’s an interesting field to follow, it will probably be another 5-plus years before robots really start to scale massively. So, investing in this space is akin to investing in AI in 2015-2016 or so, with the big J-curve in terms of demand only coming 7 years later. We think this is a good framework of how to think about the space as it likely will still take a while before the really big returns come for the winners. Next, we will review a number of the best positioned names in humanoid robotics and their supply chains, and we’ll also overview our semis outlook.

Read more

“We hosted our 3rd annual Silicon Valley AI Field Trip on 8/18-8/19, featuring AI companies, VCs and researchers from Stanford and UC Berkeley/UCSF. Model capabilities are continuing to improve, agents are progressing from assistance to workflow execution and monetization is expanding beyond seats toward consumption, transactions and outcomes. The strongest businesses combine proprietary data, domain expertise, customer context, workflow ownership, distribution, verification and permission to act. Agentic adoption should advance fastest where outputs are verifiable and responsibility for errors is clear. Open models should gain token volume across routine tasks, and frontier models should retain value where reliability justifies higher costs, supporting continued token and compute growth. AI should also expand cybersecurity demand, and physical AI will require distinct architectures and commercialization strategies. For Business & Information Services, AI should increase the value of differentiated information assets and help platforms capture more customer spending through embedded workflows. For Software, select vendors are well positioned to help customers navigate an evolving intelligence curve that segments workflows into frontier vs. non-frontier tokens and enriches models with business-specific context.”

We have a similar view and see multiple winners in this space—frontier models for the most advanced tasks, and open models for routine tasks. Companies with proprietary data will have a strong advantage—the better the quality of your data is, the better the AI’s output will be. This is relevant in both training and inference. We also continue to think that there will be obvious winners in the software space from AI, and that cybersecurity companies and data management systems such as Snowflake and MongoDB are particularly well-positioned.

CLSA tracks a wider database of AI startups and shows that the number of startups hitting key ARR milestones continues to grow rapidly:

We suspect that we’re still only at the very early start of the S-curve when it comes to AI demand. Evercore’s recent channel checks found that:

“Hyperscalers turning away customers because they don’t have enough GPUs.”

And thus AI capex continues to move higher—Nvidia sees 70% revenue growth in calendar 2027 (fiscal 2028) while the semi supply chain continues to be constrained. Nvidia’s CFO estimates that real customer demand is growing 100% year over year:

“We expect to grow revenue by approximately 70% in fiscal 2028. This is a supply-constrained outlook. With cloud industry backlog now greater than $2 trillion, CapEx by the top 5 hyperscalers is expected to reach nearly $800 billion in 2026 and $1.3 trillion in 2027. Incredibly, we are seeing demand acceleration even at our scale. Customers forecasts point to our growth, doubling next year. However, as I mentioned earlier, we expect to grow approximately 70% as we are supply constrained.”

While overall AI capex growth rates are now slowing down—36% expected for ‘27 compared to ‘26 and ‘25 which both grew around 90%—Nvidia’s comments suggest that growth rates for next year would have been almost 1.5x higher if it had not been for constraints in the supply chain.

Over the past two years, we’ve seen Nvidia’s top line growth rates decelerating while AI capex growth rates kept accelerating (chart below). However, in 2027, we see again a reversal—Nvidia will now be accelerating its growth rate again while we see a strong deceleration in AI capex growth:

One possible reason is that AI capex estimates for next year are simply too low. Another reason is that Nvidia is seeing tremendous neocloud demand from around the world—which the data above from Evercore likely isn’t properly capturing. This is Nvidia’s CFO again:

“Hyperscalers will remain a major growth driver, but non-hyperscaler growth, our AICE segment—spanning sovereign, regional NeoClouds, enterprise edge and air-gap data centers—will represent roughly half of our data center business. We don’t own a cloud ourselves, we are a neutral partner to every sovereign and NeoCloud. NeoClouds are emerging everywhere—Firebird in Armenia, Cassava Technologies across Africa, GMI Cloud in Taiwan, Yotta and Neysa in India, Firmus in Australia, YTL AI Cloud in Malaysia—pairing local land, power and operating expertise with our platform.

NeoClouds are seeing strong demand pipelines for many diverse offtakers. Rather than allocating their entire capacity to a single long-term offtake guarantee that lenders typically require to finance a data center independently, we have introduced a revenue-sharing structure. NVIDIA provides a take-or-pay commitment on a portion of the facility’s capacity, a minimum revenue guarantee that gives lenders the confidence to underwrite the project, and in exchange, we share in a portion of the NeoCloud’s revenue earned above that floor.”

We fully understand why Nvidia is doing this. The problem for Nvidia is that they don’t want a too concentrated customer base. If they would be purely selling silicon to the top four hyperscalers and the top two frontier labs, it’s likely that these big 6 will increasingly shift share to their custom ASICs over time. All are already doing this. This would make Nvidia a melting ice cube at some stage.

So, Nvidia’s answer is to set up a competing network of AI data centers comprised of neoclouds around the world, which supply AI compute to a wide number of AI startups and also innovative enterprises— like Shopify, Figma, etc. As these neoclouds often don’t have access to capital—they basically don’t have the money to build these data centers—Nvidia now provides a minimum revenue guarantee on this capacity so that neoclouds can get the necessary loans in capital markets.

We definitely think this is a rational strategy by Jensen and that he has to roll the dice here. The additional benefit is that Nvidia gets bigger upside—once from the hardware sales and a second time from recurring revenues on its infrastructure via the revenue sharing agreement. However, this works both ways. If in ‘29 or so there would be an overcapacity of AI compute, then Nvidia would see its hardware sales go into a cyclical downturn while it could also be on the hook for neoclouds that aren’t generating substantial revenues (for example, in the event that certain startups which purchased neocloud capacity go bust).

So, it’s clear that Nvidia will become an even higher beta play on AI demand, with its risk profile moving up, closer to that of a neocloud. This is great as long as there is an AI shortage, but not so great if there’s large overcapacity at some stage.

Nvidia is similarly providing guarantees for frontier AI labs to build out capacity, and from the 10Q we can see that these are substantial:

“In August 2026, we entered into guarantees with SB Energy Corp. to provide credit support on the land, power, and shell buildout at SB Energy’s PORTS Technology Campus in Pike County, Ohio, covering leases for approximately 4.25 gigawatts of IT load. The campus will exclusively host our compute under 20-year leases to OpenAI, subject to limited exceptions, with our obligation capped at $105 billion in the aggregate, subject to certain conditions including SB Energy, the lessor, satisfying applicable ready-for-service conditions. Each guarantee generally becomes effective upon commencement of the applicable lease, with corresponding guarantee amounts increasing as each of nine data centers is placed in service, which is expected to begin in fiscal year 2029. Our exposure declines as OpenAI fulfills its lease payments. Our guarantees are limited to defined portions of lease and power payments and not the full cost of the site or all of the tenant’s obligations. The guarantees terminate upon certain events, including OpenAI achieving a satisfactory credit rating or after each respective lease term has completed. We also hold an option, exercisable in our sole discretion, to provide additional credit support in phases for approximately 3.8 additional gigawatts as the site scales.”

Nvidia is guaranteeing the bulk of the cost of this site. Jensen mentioned that 1GW of Vera Rubin generates $40 billion in revenue for Nvidia, and thus the 4.25GW campus is worth about $170 billion in Nvidia hardware. With Nvidia guaranteeing $105 billion of this investment, this is about 62% of the cost.

Again, we don’t have a huge problem with this—it’s clear that both OpenAI and Anthropic are generating stellar revenue growth, with both companies combined adding $75 billion in ARR since the start of the year. And as we’re still in the early phases of token consumption, it’s obvious that AI tokens are going to become a multi-trillion industry over time. Note that OpenAI also reports revenues on a net basis, i.e. stripping out the revenue share that hyperscalers take in return for them hosting the models, while Anthropic reports revenues on a gross basis. So, OpenAI will see a huge boost to revenues by hosting their models in their own data centers.

These companies need cash to scale AI compute, and so Nvidia steps in as the guarantor so that AI labs can get the necessary loans from the banks. Again, we think this is the right move by Jensen. Nvidia makes sure that the AI labs who need to scale compute can do so, and by doing this, Nvidia cements its central position in the supply chain and continues to drive stellar revenue growth. This strong position then allows Nvidia to secure a large share of available wafer supply and advanced packaging at TSMC, memory capacity at SK Hynix and Micron, and optics at Lumentum and Coherent. In a supply constrained world, this makes it physically impossible for competition, such as AMD, to grow large amounts of share.

But again, everything hinges on AI compute not going into large oversupply. If in ‘29 or ‘30, we have a large oversupply of capacity, that massive site in Ohio isn’t worth that much. Available GPUs for which there is no demand aren’t really worth anything actually. And if OpenAI isn’t generating the revenues they expected, Nvidia might have to start paying out to cover the loans of that data center.

However, we continue to be extremely bullish on AI demand. And the key reason is that we’re using the most advanced models to write large code bases. So, being a programmer gives us a competitive advantage versus the rest of the market to really understand the progress and capabilities of the latest, most advanced models. And they truly are beyond the wildest imagination. Benchmarks don’t properly capture what’s going on. We were watching the Lex Fridman podcast yesterday with well-known programmer DHH, and his conclusions were exactly the same as what we’re seeing:

“We have AI in the pre-agentic era. I was excited about that, but it was not fundamentally rewriting the rules of the game for me. It was not completely changing how I worked. I was still chiseling code, I just had a little helper, a little sidekick- who could bounce ideas off, and I could look up this information online and so forth in a more efficient way. It was just a more efficient way to do what I was already doing, and it didn’t change the emotional connection I had to the computer.

Then we get to November 24th, 2025. Opus 4.5, to me, was the dividing line, where suddenly I gave it a couple of tasks, and I realize that the quality of the output is uncannily close to what I would’ve written. And I remember just leaning back and thinking, “What just happened?” And again, Opus 4.5 now looks like a retarded model.

I had to be in the driver’s seat. I had to tell it what I wanted, and I had to be the reviewer, the auditor of what was coming out. And then finally now, this summer, with Opus 5, Fable, and Sol, GPT Sol, and to a lesser extent, some of the open weight models, we’ve arrived at a new era where I’m not telling it where we’re going. I’m telling it the problem I have. I’m telling it the fuzzy, vague idea I have. It tells me where we’re going. It tells me which path to take, and I will still look at it because I’m a curious person and I like computers and I like the outcome of it, but I really kinda don’t have to. I’ve become optional in the part that produces the code that picks the route.

You could really get to the 100% of code written by AI and really almost– if you’re a good programmer and you have a good intuition about what’s happening behind the scenes, you can legitimately not look at the code, at least that’s been my experience. I don’t know how much prior experience with programming you need to have to kinda know that things are working correctly behind the scenes, like intuitively by observing the ripple effects, the symptoms of the system.

I’ve been working on Omarchy for the last three months, this version that just dropped a few days ago called Quattro. And almost right from the beginning I’m working on that version, the agent acceleration neared 100%, and in the last two months it has been 100%. I have not written any of the code that’s shipped in Quattro by hand. I’ve reviewed the shape of all of it. I’ve reviewed the individual lines of anything that’s critical in the model layer of the system, and I’ve not looked at a bunch of the UI code. I have not looked at a bunch of the auxiliary code, and I’ve not written any of the new functionality entirely by hand.”

It’s clear that models are now matching the capabilities of the best engineers in the world, and across more and more fields. These models are not only becoming state-of-the-art software engineers and at 100,000x the speed of the best human programmer, but also top notch mathematicians, physicists, biologists, etc. What looks like a certainty here is that there is going to be scientific progress in the coming decades at a pace we’ve never seen before. And thus, the big conclusion for us—we’re going to need a huge amount of data centers. Therefore, we continue to be very bullish on AI demand and that Nvidia is making all the right moves here.

Long term, another key attraction in the Nvidia story is that revenue per GW deployed keeps moving up. Initially, the company was making like $3-5 billion per GW. That grew to $18 billion per GW with Hopper. $25 billion with Blackwell, and now $40 billion with Vera Rubin. And after that, it’s going higher again. Note that this is positive for the entire semi supply chain—from TSMC all the way down to ASML, as this rise in the revenue per GW is driven by more advanced manufacturing processes.

So, if you install 1GW of Rubin, it gives you 30x more tokens versus Blackwell, and at a 35x lower tokens cost. All AI capacity that’s coming online in the coming years is going to be massively more productive than in the previous years, and provide much cheaper tokens.

One big concern in the market were Nvidia’s gross margins, given the explosion in memory pricing, however, the impact of this looks to be very manageable:

“As you are already aware, we are experiencing extreme pricing conditions in memory. The magnitude of the price increase has exceeded our prior expectations and are headed even higher into next year. As a result, we are resetting expectations today. For Q3, we expect GAAP and non-GAAP gross margins to be 74% plus or minus 50 basis points. We expect margins to bottom in Q4 in the 71% to 72% range before settling at 72% to 73% in fiscal year ‘28, as executed price increases take effect in Q1.”

The three key takeaways for us from Nvidia’s call were:

• Revenue growth is accelerating again and AI capex estimates for next year are likely too low.

• Nvidia is becoming a higher beta, and higher risk play, on AI demand as the company is becoming a large guarantor in the global neocloud and frontier AI buildouts.

• Memory prices will continue to move higher next year.

We continue to stay long Nvidia here. We like the acceleration story in revenues, the progress in model capabilities—which are far ahead of what many in the market are thinking—and the stellar revenue growth at the frontier labs. All while the valuation of Nvidia shares remains highly attractive at a 19x forward PE:

Read more

Broadcom has substantially lagged AI semi peers such as Marvell and AMD year-to-date, driven by fears around the company’s market share for Google’s upcoming TPUs. However, we’ve seen this movie before. Up to a few months ago, this is the exact reason why Marvell was trading at a discount—the company was ‘losing the Trainium business to Alchip’ and also ‘a shitco’.

In reality, Marvell has a comprehensive portfolio of advanced data center tech—ranging from leading edge semis and IP to optics. For example, Marvell recently closed a broad custom-silicon agreement with Google covering TPU-adjacent semis (XPU-attach). The consensus view was long Broadcom - short Marvell last year, but the reverse trade turned out to be the smart move.

Also with Broadcom, we think the market is now missing the bigger picture here. Once you start looking past the exact market share Broadcom will have in upcoming TPUs, the long term attraction is that Broadcom is the key ASIC partner for the best frontier labs—with Anthropic and OpenAI—as well as other key hyperscalers looking to compete in frontier AI—such as Google and Meta.

Our thesis is that as both Anthropic and OpenAI continue to produce stellar revenue growth in the coming years, we will gradually see more announcements around their ASIC capacity. This will help Broadcom’s multiple as the market gets increased visibility that Broadcom will indeed be a long term growth story.

However, also in the Google TPU business, JP Morgan sees continuous growth in the coming five years after meeting company management last month:

“While it was widely known that AVGO’s revenues would scale each year under its 5-year contract with Google, management incrementally confirmed that it expects to retain a majority volume share through the course of the agreement and that this was a key reason AVGO entered into the partnership. In our view, this confirms that AVGO is anchored as the primary supplier across the 5-year term, which should ease widespread investor concerns around longer-term share-loss.”

However, the real long term bull case is Broadcom’s partnerships with the leading frontier labs. JP Morgan explains:

“Management indicated that true compute demand (ex-China) is concentrated among roughly five frontier model builders that account for the vast majority of demand. While hyperscaler infrastructure can appear GPU-skewed at the aggregate level, management suggested that, when honing in specifically on leading frontier model builders, the mix could be roughly 50/50 between XPUs and GPUs next year (in line with our view that the shift from training-dominated workloads to inference monetization is pulling forward custom accelerator/XPU adoption).

The team confirmed that customers are attempting to accelerate next- generation ASIC XPU tape-out schedules and are pursuing a wider variety of chip designs. This reflects the rapid evolution of inference workloads from one-shot inference to reasoning and agentic workloads, as well as customer focus on lowering cost per token (which in our view increases the strategic importance of workload-optimized custom silicon).

Management discussed OpenAI’s first-generation XPU custom accelerator, Jalapeno, co-designed by Broadcom, and emphasized that the complexity of developing these cutting-edge XPUs is often underestimated. These products “push the envelope” across multiple design and technology dimensions, including bandwidth, SerDes, power and packaging. Broadcom expects to tape out the follow-on/next-gen OAI chip very soon.

Management pushed back on concerns that hyperscaler customer-owned tooling (COT) efforts will materially displace Broadcom’s XPU/TPU share. The key point was that modern XPUs are no longer simple single-die chips; they increasingly require multi-die configurations, advanced packaging, high-speed SerDes and deep productization expertise. Management referenced 6-die and 8-die configurations as examples of the complexity involved and suggested that internal hyperscaler efforts are likely to face significant technical hurdles.

On rising component costs, including memory, the team is aiming to keep ASIC margins intact “one way or another,” emphasizing that technology leadership is the key source of pricing power. Management also pushed back on the idea of customers consigning HBM, noting that HBM is an integral part of the XPU design and must be tightly connected to the core logic die for the product to perform properly.”

Broadcom’s head of semis showed just how massive these next-gen XPU packages are going to get:

This is Charlie Kawwas, Broadcom’s head of semis, explaining:

“This is an XPU that was built and shipped last year. As you can tell, it is made of different chiplets. Some are memory-based, some are networking-based, and some are compute. What we are building and shipping this year is this chip. As you can see, it got bigger. You can see from last year to this year, you are now getting three times more memory and the compute that is in the middle is almost three times stronger.

What ended up happening is we said for next year, we actually have to build this. So now, you have even more memory. In two years, it is almost an order of magnitude (10x). When you look at the compute, in the earlier version you had one large block, and here you have four of them. And then, this is what we are working on for 2028.

I want you to understand two things. One is the speed at which this technology is moving, and hence the innovation that it needs to go through, is extremely rapid. Two, this is how we build these chips—these are actually modular. This is why people come to us, because we actually take these platforms and pre-build them. These little chiplets that go in the chip, we pre-build them, make sure they are production-ready, work with partners—like Samsung and others—and create these memory blocks. This memory integration is one of the reasons why the prices are going higher, because we are putting so much more in them.

So, you will have 16 times what we had two years ago, 16 times in a single chip. This is ‘The Beast’—yes, actually, that is what I call it at work. We are building four of these for the largest four LLM players. It is modular.

Think of the 2025 chip as being about 1 million units a year. Think of the 2026 chip as being about 2 million units a year. But remember, the 2026 chip is three times the capacity of the 2025 chip, so that is a six-fold increase in compute capacity in one year. Then you go to the 2027 chip, which is probably 4 million to 5 million units, and this is nine times the capacity of the 2025 chip. Now you are at 40 to 50 times the scale of the first one. And then The Beast in 2028 is three times that again.

In the next couple of years and beyond—Yes, there is more than Jalapeno. There are definitely more spicy chips and a lot more XPUs that will be coming out—both from the large labs and a lot of startups. What I see is an insatiable demand in gigawatts.

For example, we have announced for one of the top five customers that this year we are putting in about 1.5 gigawatts with one of the chips I showed you. Next year, we are going to put in over 5 gigawatts, which we have already contracted with that same customer. So, we are going to go from 1.5 gigawatts to 6.5 gigawatts total. The year after that, I am pretty sure it will be over 10 gigawatts incremental. If you look at the next three years for each of these lab players, they are going to go from 1 to 2 gigawatts this year, to 5 gigawatts, to over 10 gigawatts incremental. Over that three-year journey, they will go from 2 to 5 to 10 gigawatts, which is 17 gigawatts total, starting from just 1 gigawatt today.

This means there is a massive challenge in terms of the amount of power that is needed, the amount of silicon wafers that are needed, and the amount of memory that is needed. It is actually going to pose a challenge for the rest of the non-AI businesses, which we also do at Broadcom. Remember, I have 16 franchises. Five of them are in AI, and 11 are in non-AI. AI is not just going to live in data centers—AI has to get to the edge, and these other businesses have to actually enable this edge AI. Just like the internet started from data centers and moved all the way to smart devices today. The same thing has to happen with AI.”

On The Beast, Charlie said that “we are building four of these for the largest four LLM players.” This implies to us that they are also working on an ASIC for Anthropic, and that Anthropic won’t purely rely on TPUs for their ASIC strategy. This makes sense, as you don’t want to source your hardware from a key rival long term.

From Charlie’s comments, it is clear that both 2027 and 2028 will again see strong sequential growth. Based on a recent MOU with Samsung, JP Morgan works out that also 2029-2030 should again be growth years for Broadcom:

“On Saturday, Samsung (covered by Jay Kwon) announced it had signed a memorandum of understanding (MOU) with Broadcom to expand their strategic collaboration across memory and advanced foundry technologies, supporting the next generation of AI infrastructure. We believe that Samsung is and will continue to remain Broadcom’s main supplier of high bandwidth memory (HBM DRAM), and we estimate that around 75-85% of all of Broadcom’s HBM DRAM requirements going forward will be supplied by Samsung.

The companies expect the collaboration to result in MORE than $200B of Samsung memory/ foundry products (we estimate Broadcom’s Samsung purchase mix as 90-95% HBM, 5-10% foundry wafers) over the next 5 years through 2030, and based on our calculations, we estimate that this implies > $1T of cumulative Broadcom AI revenues (ASICs + Networking) and around $40-50B of cumulative Samsung foundry-based chip sales (we believe that Broadcom will utilize Samsung for datacenter networking, broadband, and telco service provider chips).

Our estimate of >$1T in cumulative Broadcom AI revenues (2026-2030) implies a ~60%+ CAGR on Broadcom’s forward AI revenue through 2030 (>20% growth per year even after 2028) — a significant growth profile, but not a surprise given the team’s #1 market share position in AI ASIC XPUs and #1 market share position in switching/routing silicon and its Tier-1 line-up of ASIC XPU customers (Google, Meta, OpenAI, Anthropic, Asia Customer #5, Asia Customer #6, Softbank/ARM, Apple, and SambaNova).”

XPUs will continue to take share in the AI accelerator market, especially in the scenario where frontier labs such as Anthropic and OpenAI remain dominant in terms of AI revenue market share. Conversely, the more workloads move to open-source models, the better it will be for Nvidia.

When there are a large variety of models and model providers, it makes sense to optimize all of these for a key platform which is available on all clouds, i.e. Nvidia CUDA. Conversely, when there are a few frontier labs which are printing cash due to their dominant positioning in terms of AI revenue market share, this gives these labs the resources to optimize their AI infrastructure with their own XPUs. Names such as Broadcom and Marvell will see large tailwinds from this.

Note that the two scenarios are not “either-or”—both scenarios are playing out today and they will continue to play out. More simple or pre-defined background processes for which open-source and cheaper models are ‘good enough’, will continue to be deployed on these types of models. At the same time, the most advanced models will continue to win the most advanced workloads, such as writing large code bases or doing advanced engineering work.

One reason that we have the $200 per month Claude plan is that Claude can do anything very well—so, you’re writing a large code base, but Claude also understands how to translate your app into 20-30 other languages, knows how to architect the back-end system in the Google Cloud, knows how to optimize the app for marketing, etc. For example, when we tried to translate our app with GLM 5.2, ‘the Claude killer’ according to social media, it crashed. Apparently it can’t do translation. Engineers don’t like to lose hours with models that crash or can only handle a narrow subset of tasks—they like to iterate quickly and build solutions. So, even if we got GLM 5.2 for free, we still wouldn’t use it and just pay the $200 for Claude. It’s a real bargain given its capabilities.

It seems likely to us that for advanced workloads, such as engineering, companies will just continue to pay up for the most advanced models. You don’t want to work with an inferior model on a problem for 1-2 days, and then figure out it isn’t able to do it. It makes a lot more sense to purchase the best models and then get the best solutions in an instant. Engineers cost far more than tokens, and so in our view, companies will pay up for the best models to maximize the productivity of their engineers. This is where the real value is created. Not by purchasing cheaper tokens in the cloud.

We regularly see highly confident claims on social media that the “model layer will commoditize”. However, this is not how things end up in tech. Typically, there are 1-2 big winners when it comes to software. Let’s just quickly go through the winners from the last decades to illustrate. Client OS—Windows. Text editing and spreadsheets—Microsoft Office. Server OS—Linux. Browser—Chrome. Mobile OS—Apple and Android. Programming languages—Python and Javascript. Cloud—AWS and Azure. Search—Google. Code repository—Github. And you can go on like this.

In our view, when you look at revenue market shares in the LLM market 3-5 years from now, this will likely be a market where two players are fairly dominant and capture 60-70% of the market or so. Then, there will be two or more players which capture another 25% or so. Open-source and cheaper models will only capture a small revenue share of the market, given that these models are indeed commoditized. While these will process large volumes of tokens, not much money will be made in this part of the market as these will be extremely cheap tokens.

This is good news for Broadcom as these dominant AI players will allocate resources to scale their ASIC solutions. The only caveat is that—when we look at Charlie’s comments above—it’s also clear that by the end of 2028, a huge amount of AI capacity will have been brought online. The key question is whether we’ll still be in a shortage then, or whether we’ll actually have overcapacity. In case we get overcapacity in AI at some stage, this will lead to a cyclical correction in data center capex for a few years—until AI demand grows again into the available capacity. And then we’ll start a new semi upcycle again.

However, when we look at Broadcom’s valuation, we see the risk-reward as very attractive here. The company is exposed to one of the most attractive areas in AI as the key ASIC partner for the leading AI labs, while trading at a very non-aggressive valuation of 23x forward EPS:

And a valuation of 14x 2028 EPS:

Read more

We watched a podcast with Professor Robert Pape from the University of Chicago this week, where he detailed how the Iran conflict continues to worsen when it comes to natural gas and oil supplies.

There are basically three choke points around the Arabian Peninsula. One is the Strait of Hormuz, which is now almost fully closed again. Second, Iran and its proxies could also shut down the other two choke points, on the western side of Arabia, in the Red Sea.

Currently, the Saudis and Emiratis are diverting a small but significant amount of oil via these Red Sea routes:

The problem is that after the heavy fighting of March-April of this year, ‘Operation Epic Fury’, the Iranian regime hasn’t collapsed and is still, in fact, in control of the Strait of Hormuz, and with an ability to also potentially close off the other two choke points in the Red Sea. This is Professor Pape on the Mario Nawfal podcast explaining how Iran continues to escalate its demands:

“The first thing to understand here is that Iran’s demands have been expanding, not day by day, but period by period, since the war began. What you can see is in March through May, is that essentially, this is the period where Iran mostly wants peace, mostly wants a ceasefire. They have a little bit of edginess to their demands for money and reparations, but not a whole lot.

Then it transitions to the memorandum of understanding, or MOU. Once we get into the MOU, now we are in the actual bargaining phase. This is where there is negotiation between the United States and Iran on a series of these issues, and also negotiations with Iran and Oman, and Iran wants a little bit more upfront cash in this particular period.

Now we have transitioned, and I put the date as July 7th because that is when the Ever Lovely vessel was attacked. You could have put it a little bit later as well, maybe the 12th, but that would not change the fundamental analysis here. This is the period of unilateral demands. This is the period that we are in now, and I think we are going to be in for quite some time, where the demands have escalated across the board.

This can be explained fairly directly by the declining oil inventories and other buffers. This is basically the chart of the 32 countries that agreed in the middle of March to release 400 million barrels over 120 days. They are still releasing, by the way. But notice the clear trajectory here, it is going to keep going down as the days and the weeks go on because we are going to run out of the cushions. As this goes down, Iran’s leverage has been going up, and its demands have been going up.

The big thing that happened in the spring, even before the MOU, is the West came up with a set of ideas to bypass the Strait of Hormuz. I know because I have been in discussions with a lot of people from the corporate world, in April and early May, saying “No, there is not a problem here with Hormuz, we are just going to bypass it over two years.” Basically, Scott Bessent just revealed the plan, which is that in two years, Hormuz is not going to matter anymore. The reason it is not going to matter is because the Middle East is going to be riddled with pipelines.

There are already some there now, but it will be riddled with pipelines to get energy and other resources out of the Middle East through waterways that do not connect to the Persian Gulf at all. In that situation, Iran then developed, and this became public at the end of May, what the IRGC calls its ‘resistance security belt’. That is when the IRGC generals, the top leaders, started to explain that Iran did not just want peace; it wanted the security resistance belt. That was control and influence militarily from the Mediterranean, which is where the Suez Canal is, to the Red Sea, where the Bab-el-Mandeb is, and through the Gulf.

Basically, the cat got out of the bag because you cannot start talking about building all these pipelines and have this not become public knowledge. That is why Scott Bessent is publicly saying “Do not worry about us conceding on Iran, we are basically going to put Iran and the Strait of Hormuz out of business in two years.” Well, what that has done, is it has defined the goalposts for Iran’s demands. This is not a negotiation. Iran is not negotiating to get economic relief. That is a secondary issue, because if it gets relief today, that can be taken away in two years.

We are just going to put Hormuz out of business, and then we are going to be able to go back to business as usual, pounding Iran’s leaders. We are going to be able to attack Iran whenever we want to, and we will be able to come in and do more damage to Iran. So, this idea of getting economic relief is temporary at best. Power is what dictates what Iran’s security and future are going to look like, not promises by the United States.

What you are seeing here is, since essentially July 7th, where have the attacks been? This is not just a notional concept of a resistance security belt the way it was at the end of May. They are enforcing it; they are doing it. In other words, Iran has moved from survival in period 1, to ambition during the MOU, to hard power in period three. The current situation is a hard power situation where they are not looking to negotiate. They are looking to establish their security resistance belt.

What would Donald Trump have to give Iran to open Hormuz? Well, you see them in the list of demands from the Supreme National Security Council. Number one, the United States has to pull out all of its forces. So, the United States cannot stop Iran’s military control of the security resistance belt. Number two, the United States would have to give Iran money. In other words, fund Iran’s expansion and rising hegemony. By the way, there is also no discussion here at all about Iran’s nuclear program.

Notice that we are about to end the 60-day period of the MOU in just about six days, on the 17th of August. Well, there has been no extension of that MOU. What that means is there is literally not going to be any constraint whatsoever, on Iran’s nuclear programs in about six days. I think this is why we are on a collision course. This is not just a minor set of issues. We are not negotiating on the edges of how to end a war nobody wants.”

President Trump has left the US, and the world, facing a Gordian knot. Iran appears to have a massive set of demands, which look impossible to agree on—for example, giving Iran control of the Strait of Hormuz and the Red Sea, removing all US troops from the Middle East, giving Iran tons of cash, and allowing them to develop nuclear weapons. It looks like this conflict is currently in an escalation trap where things are only going to get worse.

Former Iranian ministers have already mentioned that Iran might well continue the fight into ‘29 and then negotiate with the next US President. That said, reports are mixed—Iran’s own president and central bank chief reportedly told the Supreme Leader the blockade was crippling the economy, and senior negotiators are openly arguing in public that Tehran needs sanctions relief.

The current US plan is a new blockade in the Strait of Hormuz, with the goal of cutting Iran off from the world economy. Secretary of Defense Pete Hegseth has announced that the US Navy intends to maintain this ‘steel wall blockade’ indefinitely by continuously rotating ships into the region. And US Central Command (CENTCOM) recently established ‘Task Force Falcon Strike’ to enforce the blockade using unmanned drone systems in the air, on the surface, and underwater.

The problem is that if this doesn’t get resolved in the coming months, which now looks likely, oil and natural gas prices will start spiking at some stage—either later this year, or early next year. This will likely force the Fed to raise interest rates, which can be a negative factor weighing on equity markets, especially given that valuations now are reasonably high. In case the Fed doesn’t raise rates and decides to let inflation run, well, it’s probably a good idea to buy some gold as well, which is still the best hedge against high inflation.

However, our view is that Kevin Warsh would raise rates under that scenario. He’s made it very clear that he will bring inflation down to 2%, and so we’re not holding any gold at this stage:

“There is no soft inflation target. There is no soft implicit target, not on this committee’s watch. There’s only a target, and it’s 2%.”

Memory Outlook & The DRAM Market

Memory stocks have been selling off heavily over the last few months. The fears are that DRAM and NAND pricing are now peaking. This also isn’t helped by the fact that a lot of news outlets, and other publications, want clicks and so they will publish sensationalist headlines. One of these is that SK Hynix will give a 50% pricing discount on HBM4, JP Morgan comments:

“Multiple investors have asked whether media reports about the SKH team’s pricing HBM4 for 2027 at a 50% discount vs. competitors are accurate. While we believe investor expectations of a 100% y-y HBM ASP increase are stretched, we believe the reports of a 50% pricing discount are inaccurate.

Our relatively conservative pricing assumption (below a 40% y-y increase from 2026E) is associated with three key points:

(1) We believe it is in memory suppliers’ interest to prioritize LTA products for DDR5/LPDDR5/NAND, which are at a significant margin premium over HBM.

(2) NVDA is the largest customer for SKH, with multiple product partnerships, and the company can approach the customer from a multi-year sourcing perspective.

(3) HBM could be re-priced every year, and there is room for price adjustment again in following years (the SKH management view implies a multi- year shortage and short-term HBM contract pricing might not be the top priority if the company prioritizes three to five years’ worth of LTA volume).

If SKH successfully raises HBM ASP above our expectations, this could be a positive upside risk to our EPS projection.”

Clearly there is high dispersion in pricing estimates, Goldman reckons that SK Hynix will be able to raise HBM pricing by 100% in 2027:

“We expect HBM blended ASP for SEC (Samsung Electronics)/Hynix to reach close to US$ 2.9/Gb in 2027E, which would be roughly an 87%/100% yoy increase, respectively. We expect the like-for-like basis price increase to be roughly 60% for major products, and the remaining growth to come from the mix improvement impact.”

We agree with the bullish outlook on pricing, as AI demand growth continues to surpass all expectations. Elon Musk made the best bull case for the DRAM/HBM market on the recent SpaceX call:

“The limiting factor currently is memory. The memory output is increasing by around 20% per year. Now normally, that would be fantastically fast and amazing for any large mature industry. But, the demand is increasing by 200% a year, maybe higher. So if you’ve got demand increasing much faster than supply, then Economics 101 would suggest that the price increases. It does not decrease.”

Goldman also highlights that the HBM:DRAM trade ratio will continue to increase, which will weigh on both DRAM and HBM supply growth:

“We expect the strong HBM demand led by AI servers to continue to outpace supply, and highlight the rising difficulty in ramping the latest generation HBM. This comes from the rising complexity in using a more advanced node for both core die and base die, as well as moving to a higher layer HBM stack, which results in a lower production yield. The higher trade ratio between HBM and conventional DRAM also leads to incrementally less supply of HBM as we move forward to the new generation of HBM. As such, we expect 2027 Supply/Demand to be even tighter compared to this year, which should help drive a favorable pricing environment.”

This is a nice chart from HSBC and shows that while server DRAM pricing has dramatically increased over the last three quarters due to the supply shortage, HBM pricing on the other hand has been flat as contracts are typically negotiated on an annual basis:

We can also see that $/GB pricing of conventional DRAM is now exceeding that of HBM. This should suggest still a lot of upside on HBM pricing in the coming year. The modeling from HSBC looks reasonable, in our view, especially as supply shortages continue to increase. This is Micron’s CBO at the recent Keybanc conference describing the industry landscape:

“Based on these increased signals of demand from our customers, we now expect that 2027 calendar year will be even tighter than 2026 because the growth in demand that is taking shape is faster than the growth in supply for calendar 2027 on an industry basis. On top of that, we have this issue of HBM growing very significantly. And HBM is needed in these AI systems because in a lot of the workloads, the processor—whether it’s a GPU or an ASIC or a CPU—is sitting idle for 50% of the time, because it is waiting for data from the DRAM. And that is a huge underutilization of an important asset.

And so you need to really get a much higher performance memory bandwidth and a much higher capacity of memory in the system, so that AI can be deployed on an efficient and scalable basis. There is so much demand coming from that. And this HBM increase consequently is creating this 3:1 trade ratio that we have mentioned in the past—where to produce 100 bits of HBM, we have to reduce 300 bits of DDR supply,because there is that 3:1 trade ratio between HBM 3E and DDR.

And when you go to HBM4 and HBM4E, by the time you get to 4E, that trade ratio has worsened to closer to 4:1. So all of this growth in HBM pressures the supply that is left for everything else, which means that the wafer supply has to increase dramatically, and that’s just not easy to do. It takes a long time because the whole industry got itself into a state where most of that expansion was needed to be done in greenfield expansion, which means you go to an empty space, where there’s nothing but trees, and you have to create an entire massive fab cluster there.

So that’s a very challenging endeavor. And while our customers have to build data centers and secure power and real estate; building a data center, no offense to our customers, is considerably easier than building a sophisticated leading-edge technology front-end fab. It’s one of the most complex engineering projects in the world from a construction perspective. And so it takes a long time for these fabs to come online, and this leading-edge technology that gets deployed takes a long time to ramp.

And so because of all these reasons, we don’t have line of sight as to when that supply vector will intersect the demand vector, which continues to escalate with every passing year. And they’re all telling us that the #1 constraint they have today is DRAM. The aggregate demand across all of the segments is at much higher levels than it has been.

Our customers are telling us that despite the fact that the prices are at very high levels, that they’re eager to get more supply. And it’s not just focused on the data center, it’s across different parts of the market. As we engage with customers on these longer-term agreements and near-term supply, their constant feedback is that they are not happy about the extent of volumes we are making available to them. They are signing up in these agreements, but they feel like they’re leaving considerable opportunity on the table across market segments.

Of course, it’s most acute in the data center where quite often, we are not able to meet any more than half of the demand our customers have. And so when we look at all of that, these are like multiyear signals that we’re getting from customers.”

Subscribe now

Another concern in the market is despecking—if customers are reducing DRAM content, some in the market reckon that DRAM is not that important after all. However, when you can’t get enough DRAM to ship your products, it’s logical that you reduce the DRAM content to whatever is available. Micron mentioned that they’re only able to meet half of DRAM demand from the data center market. Micron’s CBO explains despecking:

“In our interactions with our customers, it’s not really driven by pricing on the server side as much as it is driven by just lack of adequate supply. Really, our customers are struggling to get their hands on adequate DRAM supply because the extent of supply that is available to them is not going to enable them to ship the units that they need to ship. Consequently, they are focused on balancing how much capacity of DRAM to put in the system with the number of units that they want to ship. And so that is where the modulation of the average capacity in the system is happening, whether it’s in servers or some consumer product.

Despite all of that, when you multiply the units with the average capacity that our customers are planning, the aggregate demand coming from even those new levels of average capacities is rising despite that. Also, there is this phenomena about system performance getting impacted if there isn’t adequate memory capacity. And that kind of change is going to further reduce the potential utilization that the processor is going to have. That means that there is tremendous opportunity to improve the system performance simply by increasing the amount of memory in the system. Once the system is configured and qualified with a certain amount of capacity, increasing the capacity of memory in the system is an easier lift for our customers, and they would be able to introduce a higher-performing SKU and with a higher level of capacity when they see the additional supply becoming available.

Our customers also find that as they make AI systems do more reasoning and get into agentic modes, the context windows lengthen. That requires more DRAM and it ends up spilling over into NAND flash. And so we continue to see an environment where the capacity of DRAM is heavily stressed in these next-generation AI systems.

So, there is all this latent demand out there that comes from being able to increase the average capacities and create SKUs that are higher performing, able to have longer context windows, and much better capability from an overall system performance perspective. But again, it’s not clear on a multiyear time frame when that additional supply will be available.”

If more DRAM would be available currently, customers would be doing the reverse and upgrading the DRAM content of their AI servers. The problem is that the DRAM just isn’t there, and so this creates huge potential for additional demand as next-gen AI workloads scale up.

Read more

We’re always amazed how many market commentators look at these rankings and treat them like they’re the gospel:

Currently, we do all our coding with Claude inside VS Code—the popular IDE from Microsoft—and occasionally, we try out some new models for evaluation. So, one month ago, when GLM was all the rage on social media, we once again decided to do our own test to see if we can actually use this extremely cheap model.

So, we took a task which Claude was doing for us and then we sent it to three additional APIs—GLM 5.2 Max in Openrouter, the most advanced Nemotron model from Nvidia in Openrouter, and Gemini 3.1 Pro in GCP’s API. Then, we let ChatGPT evaluate the output as a neutral arbiter—Claude Opus scored 8.5/10, Gemini Pro scored 8.0, Nemotron scored 6.5, and GLM crashed so it got a score of 0/10.

This is basically the reason why AI spend—where consumers and enterprises are actually spending their dollars—is currently close to a duopoly between Anthropic and OpenAI:

Palantir made a similar conclusion, and noted to treat all the benchmaxing which is happening with a healthy dose of skepticism, this is the company’s CTO:

“We’re excited for a new era, not one of benchmaxing but of benchmaking. The assumption that the frontier is actually the best performing is just not borne out in practice. Within 24 hours of bringing Nemotron Ultra into our stacks, we found 5 production tasks where a standard Nemotron Ultra model without post-training beat frontier models. That shouldn’t even be possible. The benchmarks are right for what the benchmark is measuring, but those are not the tasks my customers were trying to solve. And so, moving this to an empirical basis is, is how we’re going to accelerate the realization of tokens to real economic value.

This underscores that a handful of common benchmarks can be gamed. That era, benchmaxing has ended. We’ve been beholden to a small number of benchmarks that people have been designing models to and then releasing models saying like, “look how well it does on this benchmark”. But the benchmark has actually almost nothing to do with your business. So how do you figure out how to make the benchmark that represents your reality, what you’re trying to succeed at, what you’re trying to get better at and then see what model makes sense?”

So, we haven’t been able to replicate Palantir’s results with Nemotron far outperforming Claude, but naturally, Palantir is also talking their own book. They want open source models to win so that new solutions from Anthropic and OpenAI won’t be a threat to their platform.

We are positive on open-source models taking share for simple or predefined workloads—e.g. an AI model analyzing a customer invoice, classifying which comments are appropriate for a social media app, analyzing images, customer support, etc.—however, for the really advanced engineering work such as writing large code bases, this will continue to be a consolidated market in our view. The reason is simple: only the largest players will have the capital to purchase and create the largest data sets—including paying professionals to create highly customized and advanced data sets such as investment bankers building financial models, lawyers explaining how to argue a case or select a jury, etc.

Subscribe now

AI is basically a scale game where both the quality and quantity of the data the model is being trained on are two of the most crucial factors when it comes to model performance. Therefore, the thesis that the model layer in the AI stack will fully commoditize will continue to be wrong. Bears on frontier models also cite this token usage chart from OpenRouter, arguing that open-source is clearly winning:

However, it doesn’t make sense to use frontier AI models via OpenRouter as you would have to pay a premium. So, everyone either connects to the Anthropic and OpenAI APIs themselves, or uses the big cloud platforms such as Amazon Bedrock. The OpenRouter chart is only useful to see directionally how fast token usage is growing, and which open-source models are being used. However, to draw any conclusions on frontier model usage, its value is basically zero.

The OpenRouter chart is a reason to take Palantir’s comments on Nemotron Ultra with a grain of salt—despite the model being fully free (Nvidia subsidizes your token consumption in return for you allowing your data to be used in training), people just pay for the other models instead.

Another flaw is that market commentators typically only look at the cost per token when assessing the economics of each model, however, one should also take into account the opportunity cost. Every time you give coding work for example to non-frontier models, the risk increases that it will do something wrong and you just lose a lot of time. A model which makes you lose hours because the code has bugs, or even worse, a bad fundamental design, has a high opportunity cost, no matter how cheap the tokens are.

Token cost is a rounding error against engineer time. A senior dev costs somewhere around $100–150 per hour, whereas an agentic day might burn $30–80 in tokens. So the frontier model only has to buy back 30 minutes a day to pay for itself entirely.

We’re on the $200 per month plan with Anthropic, and it definitely feels like a steal. The value of the plan is like having a small engineering team under your control at any time which delivers high-quality work, and extremely rapidly. We will occasionally continue to try other models—for example, we’ll try out Grok 5.0 Build when the new model releases, and the next Gemini Pro, however, we think that frontier LLMs will be a good business as there won’t be much competition. In addition, engineers also get used to working with a particular model, creating a stickiness effect like in software.

Currently, we simply code everything with Claude, and might consult Gemini Pro if Claude proposes something that doesn’t look like a great solution. Recently, in one of our apps, Claude proposed writing a very cumbersome piece of code, and when we gave the problem to Gemini, it came up with very elegant and simple code to implement the functionality. Claude agreed that Gemini’s code was much better.

Google—Taking Some Profits

As discussed above, Gemini Pro does feel like a smart model to us and we regularly consult it when Claude gets stuck, and occasionally, it comes up with very smart solutions. However, as we’ve been using both Anthropic’s and Google’s products extensively over the last 6-9 months, it does feel to us that Anthropic is massively taking the lead.

A key problem is that Gemini Pro just regularly crashes, whereas Claude is extremely stable. We can work with it on a large code base in long sessions. Claude also just gives extremely robust results—it doesn’t touch anything it isn’t supposed to touch, and just implements the changes you want.

Another key problem with Google is that Search is still 50% of revenues (and more of profits), and this business is likely going to get disrupted. At the same time, Google hasn’t released a new model in ages, and now also some of the best people in AI have just left the company.

The parts of the business we like are GCP, Youtube, and Android; but these are only around 40% of revenues combined. Given the disruption in Search, at 27x forward PE, this is an expensive name:

We’ve kept half of the position for the moment on the hope they’ll release a cool model soon, but Anthropic is just on a much faster innovation cadence. Our impression is that Anthropic is running away with this and that the gap with competition will only increase from here. Anthropic has the cash to build custom data sets, has the talent, and the compute. Google seems to be lacking in execution. We don’t want to write Google off fully, but it certainly doesn’t feel like high conviction.

Optics vs DRAM

Read more

What we liked in the Manhattan Associates earnings call—the company is a leader in warehouse management software—was a 100% conversion success from AI pilot to AI subscription. This is again a nice data point that enterprise AI adoption is progressing, while we’re still in the very early stages:

“Since our launch in Q1, Active Agents have progressed from an early adopter program to now touching over 10% of our active installed base, either through a pilot or a subscription. And while it’s still early, so far, we have experienced 100% conversion success from AI pilot to AI subscription.”

The bear case on software is that enterprises will code complicated software such as warehouse management and supply chain management systems themselves. However, we think that the market is underestimating the expertise that this requires. Vibe coding a Python app to run on your PC is very different from coding an advanced SaaS that can scale in the cloud to handle Walmart’s volumes, while being 100% secure and reliable.

For example, Manhattan also talked about how customers need the company’s forward deployed engineers to implement Manhattan’s agentic capabilities. So, the idea that enterprises will build all these systems and agentic capabilities from scratch themselves looks unrealistic to us. As we’ve predicted over the last year, bookings and top line growth for quality software names are actually accelerating:

“Some examples of these investments include building seamless agentic AI capabilities driven by Manhattan forward-deployed engineers. Three consecutive quarters of record bookings has given us confidence that our go-to-market approach is working. So regarding some of the specifics on our Q2 bookings, sales to existing customers have accelerated. And in Q2, conversions from on-prem to Manhattan Active represented over 40% of our new cloud bookings. Renewals continue to be in line with our full year plan and net new logos represented over 25% of new cloud bookings in Q2, while our win rate metric remained consistent above 70%.”

So, the consensus thesis in the market that vibe coding is going to eat SaaS, there is no evidence yet. Obviously, Claude is a state-of-the-art coder and our productivity is up 50x or so with Fable. What could be a one-two weeks coding project, Fable can code the whole thing in half an hour or an hour or so. However, this is very different from coding a warehouse management software that can scale in the cloud and run Walmart’s warehouses.

Manhattan ARR per customer is also only around $1 million. Basically, if a company would hire a small engineering and support team to replicate the system with Fable, the cost is still way more than what you’re paying Manhattan. So, the idea that companies are going to code supply chain management software themselves doesn’t make economic sense at this stage. It would only work if Fable can do everything—code it all out and then provide support to end-users, with maybe one or two engineers for oversight. Still, we don’t think it’s worth the risk even in that scenario as switching warehouse management systems are complicated multi-year projects that are a massive risk to the business and P&L. Are you really going to risk your business to save $1 million by moving to your own, unproven, vibe coded supply chain management software?

We think that 90% plus of existing customers will decide to stay on Manhattan, and perhaps 5-10% of the rare, top enterprises will write these types of systems themselves with an engineering team assisted by Fable. So, we think that the call the market is making will be the exception rather than the rule. At the same time, automating workloads for customers with agentic capabilities represents a new opportunity for software. This is also what the recent Bernstein survey found:

“CIOs increasingly expect AI workloads to be cloud-based and prefer packaged solutions from established software vendors over building in-house applications. Importantly, most do not expect AI to replace enterprise software, reduce long-term IT budgets, or drive a shift from software spending toward hardware investments.

Although GenAI is a leading investment area, CIOs do not expect to increase spending on LLM vendors such as OpenAI and Anthropic, reinforcing the view that enterprises prefer consuming AI through established software platforms rather than building capabilities in-house.”

Obviously, Claude is a phenomenal model but enterprises will prefer to fine-tune open source models themselves for their particular workloads, while also simply levering agentic capabilities within existing software systems. There is still a big engineering skillset needed to build all these agentic capabilities, and most companies simply don’t have that talent. So, Manhattan is doing well and this business will continue to do well in our view.

The market has started to realize that selling these shares down to a 20x PE didn’t make much sense as bookings and revenue are accelerating:

Subscribe now

Data Management Systems as a Sweet Spot

Manhattan Associates is not the only software name seeing a tailwind from AI. As Needham’s customers checks confirm, data management systems in general are seeing an acceleration in spend:

“Last week, our customer checks indicated Enterprises are definitively using PostgreSQL for Agentic-based, external-facing applications. However, organizations are expressing a more nuanced view of the Database market (compared to investors’ black or white extremes), with the understanding that we are in a polyglot environment. Each database is capable of delivering a portion of the Agentic AI stack.

MongoDB was positively reviewed as capturing a majority of net-new, Agentic-first applications; with customers expecting a yr-yr acceleration in CY26 spend. MongoDB is favored for its JSON document model; offering a flexible schema, promoting Developer velocity, and well-suited for persistent memory use-cases (where the current state of the workload requires memory beyond a single session - like multi-step Enterprise workflows, as an example).”

So, enterprises are increasing their use of various data systems to supply AI agents with the data they need. For smaller applications where engineers are comfortable with SQL, Postgres is a no-brainer. However, if you need vast, unlimited data scaling capabilities and with flexible schemas, MongoDB is the top choice. And if you need advanced analytics capabilities, Databricks and Snowflake are the top choices.

Jefferies notes that Databricks raised more capital as they’re short on GPUs—once again a good indicator of enterprise AI demand. One of our readers had the opportunity to invest in Databricks at $70 billion last year and we told him he’d easily double his money. It’s going to be much more than that. From Jefferies:

“Databricks is raising $3B, led by Coatue, at a $188B valuation – a 40% step-up from the $134B mark set in Dec 2025. Assuming a 65% revenue CAGR from FY26 to FY28, this implies ~17x FY28 rev. SNOW currently trades at just ~13x FY28 rev on a $100B EV. We see SNOW as a key beneficiary of this read-through, as a rising tide in data cloud demand lifts multiple boats; even a discount at 15x would imply a $310 stock and a ~$115B EV. Reiterate Buy.

Databricks’ CEO recently told CNBC that surging demand for both proprietary and open-source models has left the company running out of GPUs across multiple regions. “We nearly exhausted our GPU capacity in Asia, and demand is rising in countries including Japan, South Korea, the United States, and India. We therefore need to acquire a large number of additional GPUs, which requires significant funding. That demand was what triggered our latest fundraising round: we were inundated with customer requests and needed more GPU capacity”.

We continue to believe the leading data analytics vendors, Databricks and SNOW, are best positioned to help organizations make sense of their business data and apply AI to run analytical workflows faster and more efficiently, and we expect momentum to build across the board. Databricks’ Genie and SNOW’s CoCo/CoWork are both gaining traction, with early adoption metrics inflecting. At the same time, we view the two companies’ aggressive investment in AI-driven analytics as making the competitive landscape incrementally tougher for PLTR. In particular, Databricks’ Genie Ontology adds an enterprise context layer and knowledge graph that could encroach on PLTR’s core value proposition.”

We disagree on the final point and don’t think that Snowflake and Databricks are really threats for Palantir. Palantir is an operational decision layer for lower tech-savvy end-users, whereas Databricks is basically an analytics platform for data scientists. At the same time, Snowflake is heavily focused on business analysts with a SQL background. So, we think that all three will continue to see strong growth, and clearly Databricks and Snowflake are hugely benefitting as enterprises are consolidating their data in the cloud, and increasingly using AI workloads within those platforms.

So, we can see that Databricks has built a UI for lower tech-savvy users to compete more head to head with Palantir (image below), but Palantir has been building out their platform for 10 years. So, we don’t think it will be that easy to actually start winning deals against them.

It’s something to keep an eye on though, as this is clearly a UI that all employees in an organization can use to derive business insights from and then make key decisions, while being powered underneath by the powerful Databricks analytics platform.

We like both MongoDB and Snowflake. We continue to hold Snowflake and wouldn’t mind owning Mongo, as both these businesses will see tailwinds from enterprise AI adoption.

Palantir is starting to look interesting as well—EPS will double this year for a company that’s now trading on a 77x forward PE. This is a name we could pick up if the market keeps selling it.

Next, we’ll go through more investment ideas, data points, and insights from across the AI space—including semis, software, and other names. We’re only one-third through this post.

Read more

We’ve written extensively on EDA (electronic design automation) software in the past. For newcomers on the topic, a recent expert call with a VP at Analog Devices gave a good overview of the space:

“EDA varies if you’re doing digital or analog design. Let me start with digital because that’s where a lot of companies are focused. In doing digital design, you start with a specification of what you want to build. That specification is normally in some form of natural language. That specification is then converted into the front-end logic design, where engineers use the different logic components to implement that design. After this design phase is the RTL generation phase.

RTL is actually the code to implement the design—this can be in something like Verilog (a hardware description language used to model electronic systems). That RTL then goes through a verification step. Essentially, we test that logic to make sure it’s working properly. After that phase is complete, the back-end or physical design phase happens, where you physically map the logic design to the chip area that you want to implement. Once the physical design is complete, another round of testing is done to make sure all the logic, all the timing is all still working.

Only after that final phase of testing the physical design is the design ready for tape-out in manufacturing. Those are the high-level steps to build a microcontroller or some form of digital design. Most semiconductor companies will leverage a lot of software tools to help them through these different phases. The two most common partners to work with on this whole workflow are companies like Cadence and Synopsys, and Siemens as a third. They make this much easier by providing an integrated workflow to accomplish these different phases.

We use both companies. Cadence we leverage for a lot of our analog design and Synopsys for a lot of digital design. We’re leveraging their flows pretty much end-to-end—all the way from the front-end tooling for RTL creation, their compilers, their DV (Design Verification) test benches, all of it. These companies provide tools like simulators as well, so that once you start building your design, they provide a lot of capability to simulate that. That’s part of the testing phase.

I’d say that the most painful step, or the one that probably takes the longest and is the riskiest, is design verification (DV). It’s essentially testing the design and like most technology—whether it’s software or hardware—getting all the corner cases, generating what are called behavioral assertions, this is probably the hardest part of any design. We struggle with this constantly to get the proper test coverage. The RTL development simulation is not that complicated compared to the DV test bench at the end. There’s also going to be some challenges on the physical design, the place and route, and all the timing closure, getting it physically to fit within your constraints.

We also leverage different types of IP from these companies, so we license parts of the design. Those individual IPs have existing test benches, that already exists. What becomes more bespoke is when you take their IP and combine them with a bunch of our internally developed IPs, and then you integrate all that into a SoC. The corner cases at the SoC-level are new and you have to create new test cases for each individual device. Think of it as the components are individually tested, but when you integrate them it becomes a challenge.”

Although Cadence was historically stronger in analog, they’ve been closing the gap in digital semi design with Synopsys. This was another good recent expert call with a semiconductor veteran who’s now on the board of Baya Systems (a startup in Silicon Valley working on chiplet connectivity):

“I did start-ups, four of them. Two of them ended up ultimately becoming a part of Synopsys, and then one of them IPOd. You can’t think of Cadence without thinking about Synopsys because it’s a duopoly. If you are designing chips, you are going to use both companies. They have similar histories of how they were started. Some of the same people went back and forth between the companies. Think of them as utilities or tools that, as semiconductor design was starting to use PCs and mini computers, these little tool companies started merging to become what is now Cadence and Synopsys. Combined, they have 80%+ of the market in most sub-segments of EDA. Even semiconductor IP, they’re very big there.

As far as the business model, both of these companies, when you’re dealing with the larger companies, they have a license that might be three years or five years. For smaller companies, it might be something that’s more token based, you generally don’t get all-you-can-eat for all your engineering employees to use the tools. You get a more restrictive license—and probably a higher margin license for the EDA companies—than what the big companies have.

The EDA companies work together with the fabs to create PDKs, process development kits. When TSMC, or Samsung, or Intel, come out with a new process, they work with the EDA vendors to create the kits that the end customers will use to make their designs that target a particular semiconductor process. There’s a ton of engineering work that goes into creating these PDKs. Intel has always sucked at this. Intel has gotten better, but the first time they tried to become a fab, it was impossible for anybody to use Intel’s process, except Intel, because they didn’t have a PDK.”

The EDA Crash

The EDA space has been a very attractive space for investors over the last decades. A duopoly meant strong pricing and high margins, while the recurring nature of software revenues meant that these companies weren’t exposed to the brutal cyclical swings of the semiconductor industry. At the same time, as the cost of advanced semi design exploded in the EUV era, both Cadence and Synopsys enjoyed a massive tailwind to grow their revenues with advanced semi design costs surging node-on-node:

The above tailwinds will remain valid in the foreseeable future. However, Cadence and Synopsys shares are down around 14% since Moonshot’s Kimi K3 autonomously designed a functional semiconductor chip in 48 hours, using exclusively open-source tools. The chip was a 4mm² die running at 100MHz on the freely available Nangate 45nm Open Cell Library—designed, verified, and simulated entirely without human intervention.

A few points to note here. First, the Nangate 45nm Open Cell Library isn’t a PDK and has no foundry behind it. Nangate was absorbed into Silvaco years ago—a small and low-end player in EDA—and the library is basically the “hello world” of open EDA research (constructing a “hello world” program is the typical introduction for students to any new programming language).

The problem with open-source EDA tools, however, is that these tools don’t have access to any advanced PDKs. The EDA duopoly has a large moat—not only is decades of semi design know-how ingrained into these tools, but these tools are also the only ones with access to advanced PDKs. For example, TSMC certifies tools through its Open Innovation Platform, and that list of tools is: Cadence, Synopsys, and Siemens. Then, these PDKs are also only available to TSMC’s approved customers and under strict licensing agreements. Obviously, foundries don’t give away their state-of-the-art IP.

So, what open-source EDA tools actually have access to is a much older tier of the stack—TSMC 65nm, GlobalFoundries 12nm, and Intel 16nm processes. Even if you want to design a TSMC 65nm chip, you still need a TSMC license, so it’s not open to everyone. The only fully open PDKs are very old processes, such as SkyWater 90nm and 130nm, and GlobalFoundries 180nm.

Most of the money in semis is made on the advanced nodes, and there really is no other way than using the established EDA tools if you want to design chips on those nodes. For example, TSMC requires a Synopsys PrimeTime static timing analysis before tape-out. Looking at TSMC’s revenue split, older nodes—where open source tools are competing—are a very small part of the business:

Given the semi industry’s large skew—both in terms of revenues and R&D—towards advanced nodes, we suspect that a high proportion of Cadence and Synopsys revenues are coming from the advanced nodes.

The Massive AI Opportunity

We’ve written extensively on the opportunity for EDA vendors to play a larger role in the semi design flow in the future with AI automation. Goldman recently published an in-depth report with a similar conclusion, and calculates that this represents a large opportunity for the EDA names that the market is missing:

“Consensus has viewed Cadence and Synopsys as “defensive” 12%-15% growers at best, and potential victims of AI disruption at worst. However, our new analysis suggests the shift toward custom AI silicon has exacerbated a structural shortage of chip design engineers that EDA companies are uniquely positioned to monetize with Agentic AI—an incremental opportunity we estimate at ~$3.7bn/year by 2030, which is not reflected in Street estimates and may start to be evident as early as 2H26. We believe consensus reflects the “pre-AI” version of the industry as growth is about to inflect, creating an opportunity in the stocks today.

Chip demand could drive a global shortage of ~72,000 chip designers by 2030. A key industry benchmark study (SIA/BCG, 2022) projected a ~23k shortfall in US chip designers before generative AI. Since then, hyperscalers and AI labs have launched custom silicon programs requiring at least ~2,500 designers. Meanwhile, the supply of human designers grows at just 2%-3% annually. We estimate AI design tools can add ~48k agentic engineer-equivalents by 2030, which can partially close this gap.

We size the Agentic AI revenue opportunity for EDA at ~$3.7bn annually by 2030 based on modest adoption assumptions. We estimate “Copilot-style” AI EDA tools could price at ~$25k per engineer annually, relative to the ~$90–$100k customers spend in EDA per seat. We also think autonomous agents, which both EDA vendors will launch in 2026-27, could handle up to 30% of work by 2030 at ~$50k per agent—or ~25% of the fully-loaded cost of a human design engineer.

Street numbers reflect essentially none—and we believe initial proof points could be evident within months. Consensus models both companies decelerating revenue growth from here, implying zero credit for this new Agentic layer. We estimate that this opportunity could lift Cadence’s revenue growth from ~13% to the mid-20% range (+$2.30 of EPS by 2030) and Synopsys to ~20% (+$2.20). 2H results could provide the first evidence of this growth opportunity. If agentic AI tool adoption gains traction, we believe CDNS/SNPS could re-rate in terms of faster growth rates and higher multiples.”

Is Goldman correct that AI will represent a massive opportunity for Cadence and Synopsys? Or are they missing something? Fortunately, we have a lot more insights that we came across, we’re not even halfway yet through the analysis. We’ll go through these findings next.

Read more

Our view is that we’re in the midst of a long term semis bull market driven initially by the AI data center buildout, which will be followed in the 2030s by the rise of physical AI such as humanoid robotics. TSMC’s CC Wei has a similar view:

“I believe from this day on all the way to probably 2029, 2030, the demand is very strong. Whether in between there’s a dip or not, I’m not very sure. But the trend is so robust that I believe we are witnessing a kind of a new industry. I would like to say the new AI industry, which is so common in our daily life, it’s going to affect automotive, humanoid robots, and also impact all the industry. The amount of money put in by the CSPs alone is a very important new industry to the world, so the demand will be there. And the fundamental thing about semiconductor chips, most of them are TSMC.”

While scaling laws continue to be strong in training:

Most of the growth in AI demand in the coming decades will be in inference:

Due to the combination of scaling laws and then the ramp in inference, demand is currently off-the-charts for TSMC’s leading edge nodes. The company showed this nice chart at their recent symposium which is indicating a steep increase once again in AI accelerator demand this year:

In the years ahead, Wei has to estimate where the real demand is going to be. This is how he goes about it:

“Because all the customers are very aggressive, and that’s the CEO’s job. CEO got to be aggressive. They give me the number of their demand, and I believe they try their best to tell me the truth. So, I put all together, and all the truths together is not a truth. Mark down that word. So, yes, we do a very careful judgment. May not be correct, but we did carefully because this is big money, right? So, we are checking the AI data centers progress, the building, the location, the demand, the racks, we’re checking all that to make sure that TSMC chips will not be put in inventory. This year, we say we increased the CapEx from $52 billion to $56 billion, now $60 billion to $64 billion. And you bet, that will continue to increase. But, we do it carefully.”

Even with the newly announced capex increase, TSMC’s capex-to-revenues ratio remains at fairly modest levels:

Thus, while TSMC is making new heights in capex, there actually remains plenty of room for growth in the coming years:

Despite TSMC being capacity constrained, the company won’t be doing any exuberant price increases like we’ve seen in Memory. Wei took a few jabs at the Memory players:

“Yes, one of my competitors in South Korea, they make a huge amount of money, and I’m jealous about it. What is the wafer pricing strategy for TSMC, and what kind of gross margin we should have? The higher, the better, of course. But we are a partner, meaning that our customers got to be successful. I don’t want to squeeze them out from the market. And besides, we are a very trustable company with our customers. So we don’t suddenly increase our price by 4x or 5x. Your customer cannot survive for that kind of a price increase. So we earn our value, and we make sure that our profit, our gross margin is enough for a long-term sustaining expansion. That’s to the benefit of my customer and TSMC also. That’s our philosophy. So yes, I’m really jealous about the memory companies, 86% gross margin. I would be happy with that. But, anyway, we are very trustable.”

Wei also knows that the risk with massive price increases is that it will attract new competition over time. While manufacturing DRAM is an advanced set of processes, a key reason that the industry is highly consolidated is that it’s both highly capital intensive and, historically, extremely cyclical. While it has been very pleasant to be a shareholder in these names over the last year, the longer pricing remains at these levels, the more competition will get oxygen as well.

We can already see this with CXMT, China’s DRAM manufacturer, which is seeing a boom in revenues as well given that they’re a price follower. This is also a key reason why the SpaceX-Tesla Terafab will focus on manufacturing DRAM. While neither of these two can really pose credible competition in the coming 3-5 years, exuberant memory margins will attract capital to fund more competition in the industry, and these new players can become more credible competition in the 2030s. While CXMT obviously doesn’t have EUV, the more cash that the Chinese fabs can generate, the more cash they can deploy into the Chinese tooling ecosystem for R&D purposes as well.

Chinese tool manufacturers have gradually been climbing the technology curve over the last few decades, and players such as Naura, AMEC, and ACMR are increasingly manufacturing more impressive tooling. While China is currently far behind in lithography and metrology—the more cash the Chinese chip ecosystem has, the more likely it is that China can pose a threat in advanced semi manufacturing 5-10 years from now. While DRAM manufacturers are optimizing for near term profitability, long term this can come back to bite them. NAND is even worse as this industry is much more commoditized with five larger players plus now also China’s YMTC, which has managed to grab a 13% global market share. So, NAND has already effectively become a six player market.

Advanced semi manufacturing is increasingly considered to be crucial for national security. This is why the Japanese government in recent years has been funding the country’s return to leading edge logic manufacturing with Rapidus. With the current astronomical margins in DRAM, we wouldn’t be surprised to see similar announcements in DRAM, with geopolitical powers such as Japan and Europe potentially calculating that they will need to secure supply in the form of a national DRAM champion. Both these regions were still strong in semi manufacturing only a few decades ago, but were eventually disrupted by Korean and Taiwanese rivals. Europe’s last major DRAM manufacturer, Qimonda, only went under in 2009 following the financial crisis.

Also from a geopolitical perspective, dramatic memory price increases are narrowing the advantage the US has over China, as these will slow down the data center buildout, and raise token pricing. As already mentioned, huge price increases are also making Chinese fabs cash rich, which will benefit the entire Chinese semi supply chain due to higher R&D investments.

Subscribe now

So, TSMC’s strategy is really to be the trusted technology partner for all leading edge chip designers, prioritizing long term relationships over short term profits. This is really the beauty of TSMC’s business—while silicon design is much more competitive, TSMC simply takes care of all manufacturing, whether it’s CPUs vs GPUs, or even all the various types of CPUs. Wei explains the company’s strong position:

“The AI market continues to be very dynamic. The emergence of Agentic AI is leading to a resurgence in the role of CPUs in AI data centers, which drives more silicon demand in addition to AI accelerators. We believe this is positive for TSMC as no matter what CPU approach is taken, whether it’s x86, ARM-based, or RISC-V architecture, they are almost all TSMC’s customers. We are already collaborating closely with our CPU customers and working to support them with the most advanced technologies and necessary capacity.”

Customer interest in N2, TSMC’s latest node, is strong—with TSMC having more than 20 customers and over 70 in the pipeline:

TSMC is planning a steep ramp of N2 capacity in the coming years:

Both Fab 20 in Hsinchu and Fab 22 in Kaohsiung are ramping strongly, with capacity this year already surpassing the N3 ramp in its first year:

At the same time, the company continues to grow N3 capacity, with tools now being relocated from N5 to N3 fabs:

Then, in two years from now, the A14 ramp will start, which will again be followed by more advanced and more specialized variants of this node:

A14 offers again higher speed with higher chip density:

Wei mentioned that the lead time for a new technology such as A14 is 5-7 years—the technology has to be developed, tested, improved, and then capacity has to be ramped.

Next, we will go through much more findings and insights on TSMC, leading edge semis, and what’s next for both. Finally, we’ll also review the current semis sell-off, which names are starting to look attractive here, and which names we’d still avoid.

Read more

MEMS used to be a fairly dull corner of the semiconductor industry and so historically, we’ve paid very little attention to this market. However, this is changing as optical MEMS are the backbone of one of the most cutting edge markets in the AI data center with optical circuit switching (OCS). In addition, MEMS will see long term growth from physical and edge AI in end-markets such as robotics, drones, satellites, IOT, etc.

In the AI data center, servers typically rely on layers of electronic ethernet switches to communicate with each other. The problem is that converting signals each time from the optical domain to the electrical one, and then back into optics—i.e. optical-electrical-optical conversion—takes a lot of power, and so it makes sense to keep that signal fully in optics in the first place. This is where OCS comes in.

At the heart of OCS are MEMS-based optical switches, which are basically movable micromirrors. These have a very simple function—redirect an incoming optical signal from one channel into another to reroute traffic. Yole illustrates what this looks like:

Each input-side mirror can be oriented toward a corresponding output-side mirror, effectively creating an MxN optical switching cell. The win is that OCS cuts power consumption by up to 40% to 50% compared to equivalent electrical switches. It also introduces near-zero latency because light moves through the switch at the speed of light in glass/air.

There’s one big drawback however—switching speed. MEMS mirrors are mechanical devices. Moving a physical mirror takes time—typically between 1 and 10 milliseconds, compared to traditional electronic switches that can route data in nanoseconds. Because of this millisecond-scale switching speed, OCS cannot be used for dynamic, packet-by-packet routing. Instead, it is used for a particular topology reconfiguration, for example, for re-wiring the data center fabric to connect huge pools of XPUs for a training run that will last hours or days.

Yole expects the optical MEMS market to grow at a 23% CAGR—OCS is the market in the light blue bubbles:

Silex Microsystems & MEMS Manufacturing

Silex Microsystems is the largest MEMS foundry in the world and was recently IPO’d in Stockholm. The company manufactures the optical MEMS for Google’s OCS. From the founder:

“I founded the Company together with four other PhD students from KTH Royal Institute of Technology in Sweden, with the aim of creating a pure-play MEMS foundry fully dedicated to manufacturing MEMS based on customers’ designs. The Company swiftly outgrew its research laboratory origins, shifted to a new production facility in Järfälla, Sweden, and in 2009, transformed the MEMS industry by introducing the world’s first 200-millimetre (8 inch) wafer MEMS manufacturing fab. We have worked hard to drive the technological forefront of MEMS manufacturing since then. Today, Silex is the leading pure-play MEMS foundry globally. We have built a large customer base and long-term partnerships with many of the world’s leading technology companies.”

MEMS stands for Micro-Electro-Mechanical Systems. In simple terms, MEMS are microscopic machines which combine tiny mechanical elements such as sensors, actuators, and electronics on a silicon chip. MEMS devices usually range in size from a few micrometers to millimeters. To put that in perspective—a human hair is about 75 micrometers wide. Many MEMS components—like tiny gears, mirrors, or beams—are often just 1 to 100 micrometers in size.

MEMS are used in loads of end applications. For example, in a smartphone, MEMS detect when you rotate your phone from portrait to landscape, and they can track the number of steps you take. The same happens in Nintendo Switch controllers—MEMS motion sensors track the user’s movements to then translate them into the game.

Before MEMS, these types of devices were bulky pieces of hardware made of springs, weights and wires. However, as MEMS use the same manufacturing process as in semiconductors—i.e. photolithography, etch, deposition etc.—this allows MEMS to be mass-produced onto silicon wafers for cents on the dollar.

The MEMS universe is categorized by what the tiny mechanical parts actually do:

• Optical MEMS: These use micro-mirrors, lenses, or gratings to manipulate light. Optical MEMS route optical signals through fiber-optic networks and drive the spinning mirrors inside lidar sensors for self-driving cars.

• RF-MEMS handle high-frequency wireless signals such as in 5G infrastructure, smartphones, and radar systems.

• Inertial & Acoustic MEMS measure physical motion, forces or sound waves. These track movement, rotation and sound in consumer electronics.

• Microfluidic & Bio-MEMS feature microscopic channels, pumps and valves to move tiny droplets of liquid with absolute precision. These are used in DNA sequencing, “lab-on-a-chip” medical diagnostics and precise drug delivery systems such as insulin pumps.

• Power & Actuator MEMS are designed to physically move or harvest energy at the micro-scale. Silex utilizes a specialized material called PZT (Lead Zirconate Titanate) in its manufacturing. This is a piezoelectric material—meaning when you apply electricity to it, it physically bends.

Silex is focused on the manufacturing of more differentiated and higher value MEMS. They deliberately avoid commoditized, mass-scale products. The attraction of this business is that once you’re manufacturing a high value component for a customer such as Google, that customer will rarely switch as the intensive R&D to set up the manufacturing process would have to be repeated at the new foundry. However, huge customers such as Google do split their orders, e.g. 75% to Silex and 25% to a secondary foundry. Dual sourcing is common practice across the semi supply chain, with few exceptions.

According to a Japanese source, Silex also pioneered Through-Silicon Vias (TSVs) in MEMS manufacturing (Sil-Via™ and Met-Cap™). Normally, connecting a tiny mechanical sensor to a computer chip requires running delicate wires out the sides. Silex brought TSVs—which are common in semi manufacturing—to MEMS. These TSVs allow Silex to stack a MEMS sensor directly on top of a chip, making the final hardware incredibly compact, hermetically sealed, and perfectly suited for high-vacuum applications like aerospace or advanced medical tech.

The prospectus explains the deep relationship between foundries and customers to bring to market a new product:

“Pure‑play foundries serve as critical enablers for Fabless companies and design‑centric customers. Their involvement often begins at the early development stage, where repeated iteration and refinement are required to reach scale production of the final MEMS. Successful scaling from prototype to production involves coordinated work across design, modelling, process engineering, metrology, and packaging.

As fabrication techniques advance, Silex believes that, in the future, MEMS can become even smaller, more powerful, and capable of performing increasingly advanced tasks, which can drive technology innovations further, making MEMS foundational for the future of smart, connected devices. The pace of technological advancements remains high, and foundries that can adapt their processes to meet the needs of diverse applications gain a substantial competitive advantage as compared to those that are unable to do so.”

So, we also have sort of a Moore’s Law for MEMS, where these devices continue to shrink in size to enable more advanced applications. Overall, this looks to be an industry with high innovation potential and a long term growth trajectory. Silex illustrates innovation in this industry over the last decades:

MEMS are used in loads of end applications—including in high growth markets where edge AI will become important such as robotics, drones, satellites, robotaxis, IOT, etc. From the prospectus:

“Future AR display technologies are currently being served by Silex, who as a MEMS foundry is well equipped to serve the demands required to create very small form factor and complex manufacture of chips that interact with the physical domain. Silex believes it is well positioned to serve emerging device applications beyond the current defined MEMS device market, including AR display technologies that are not presently classified as MEMS devices.

The increasing adoption of MEMS across a broad range of applications has resulted in diversified demand for advanced devices. Underlying markets with rapid growth such as MedTech, AI, autonomous systems, AR and satellite communications, drive both innovation and volume in the MEMS market as new applications require novel solutions. The specialised requirements and innovative nature of these applications are expected to primarily strengthen the growth of Silex’s Focus Market.”

Subscribe now

MEMS manufacturing is still largely an IDM (integrated device manufacturer, i.e. product design and manufacturing within a single company) market:

However, foundries are gradually taking share in this market, and this trend is likely to continue in the coming decades as new technologies will need specialized MEMS design and manufacturing:

Next, we’ll detail further insights into Silex Micro and the MEMS industry, with our thoughts on investing in this space. We’re currently halfway through the analysis.

Read more

While there are many end products that make use of rare earths, permanent magnets are overwhelmingly the key application driving the current geopolitical rush. According to International Energy Agency (IEA) data, permanent magnets account for roughly 95% of the monetary value of global rare earth consumption. The key REEs for magnets are neodymium (Nd), praseodymium (Pr), dysprosium (Dy), and terbium (Tb). These four are also called the ‘magnet rare earths’.

A neodymium-based magnet is the strongest type of permanent magnet available on Earth. However, when you mine rare earths, you have to dig up all elements together. This creates a massive market imbalance—elements such as cerium and lanthanum make up the vast majority of what is physically dug out of the ground. Because these are in oversupply, they are very cheap and often trade for just a few dollars per kilogram. These low value REEs dominate volumes for low tech applications such as catalysts, glass, and polishing powders.

Pure neodymium is a soft, silvery-grey metal. On its own, pure neodymium is not a useful magnet. NdFeB (Neodymium-Iron-Boron) is the specific alloy created by combining neodymium with iron and boron. This combination locks electrons into a specific structure with high magnetic energy. As a result, geopolitically crucial end markets such as robotics, aerospace, electronics, data centers, and EVs will be dependent on NdFeB magnets:

The big problem, however, is that up until recently, all rare earth refining was done in China. For example, in 2024, the US accounted for 16% of global rare earth production, however, these rare earths were then shipped to China for refining and magnet production. Needless to say, this gives China strong leverage over the US as they can halt, for example, manufacturing of US military equipment, and also block new advanced industries such as humanoid and industrial robotics from scaling up.

Goldman shows how China in 2024 had a 59% market share in rare earth mining, but an over 90% market share in both rare earth refining and magnet production:

Goldman estimates that NdFeB magnet demand could easily triple in the coming ten years. There is clear upside to this estimate if new industries such as robotics, edge AI, AI compute in space, and EVs take off. Due to environmental and other regulatory hurdles, limited access to high-grade deposits, high capital costs, and long lead times, it’s likely that there will only be a few players in the West that can provide rare earth mining, refining, and magnet production—i.e. fully vertically integrated mine-to-magnet operators.

Subscribe now

As Western governments and big tech players such as Apple are now willing to pay premiums for non-China sourced magnets, and sign long term supply contracts, we see the long term outlook as attractive in this space. This is becoming less of a commodity market, and more of a high growth market with floor pricing guaranteeing attractive margins.

Next, we’ll go through key assets in the magnets and rare earths value chain, and we’ll review the key players involved. We’ll also discuss future TAMs, profitability, and potential returns.

Read more

We enjoyed Google DeepMind researcher Yao Shunyu’s interview on the Zhang Xiaojun podcast. Previously, Yao was at Anthropic where he helped train Claude 3.7, now he’s involved in the Gemini 3 series. We’ll give the highlights below:

“The pace of model improvement is not slowing down at all. It is hard to quantify the velocity curve because if you only look at standard benchmarks, progress will naturally appear to slow down as scores approach one hundred percent. However, the qualitative difference felt by the user does not scale linearly with benchmark gains. Going from seventy percent to seventy-five percent on a complex task can feel like a much larger leap in usability than going from fifty to sixty percent. As a researcher, my personal observation is that the models’ fundamental capacity to learn is getting stronger. Tasks that used to require massive engineering effort to teach a model can now be achieved with far less friction. The bottleneck is no longer the model’s capacity to learn, but our ability to clearly define the problem and construct high-quality data environments.

My experience is that pre-training scaling has not hit a wall, and I do not see any signs of it plateauing. When people claim the Scaling Law is reaching its limit, it is usually due to one of three things. Either they believe the physical limits of the paradigm have been reached, or they believe we have run out of usable data to scale further, or they simply have an unnoticed bug in their own implementation. In the vast majority of cases where teams hit a wall, it is due to the third reason. In this industry, fixing a single subtle bug in your scaling science or data pipeline can yield progress that far outpaces any fancy algorithmic trick. When you run into a bug, if you believe the paradigm is limited, you declare that you have hit a wall. If you believe it can be fixed, you keep digging. This is why systematic problem-solving, which both Gemini and Anthropic do exceptionally well, is so critical. You must design clean ablation experiments to isolate and test every variable.

Coding has two massive advantages that make it perfect for AI. First, it has a highly objective and well-defined reward signal. If you ask a model to write code to implement a specific feature, you can easily run tests to see if the input matches the expected output. The feedback loop is immediate and unambiguous. Second, coding has an incredible, decades-long data foundation in GitHub, which provides a massive repository of high-quality code written by excellent programmers. This allows researchers to build highly robust simulation environments. From a product perspective, the user requirements for coding are relatively singular and standardized. Unlike social media or gaming, where user tastes are highly subjective and diverse, excellent programmers generally agree on what constitutes clean, concise, structurally clear, and well-abstracted code. This consensus makes developing coding products much more straightforward. AI can write specific code blocks incredibly well, but breaking down a massive, ambiguous project into actionable, modular steps is still a human-dominated skill—at least for now.

A conservative estimate is that over ninety percent of the raw code I use is model-generated. If I am being less conservative, it is closer to ninety-nine or one hundred percent. However, I still spend a significant amount of time reviewing, editing, and verifying the logic to ensure it aligns with my exact design. The core of software engineering has shifted from writing syntax to designing system architecture, defining logic, and providing the model with the right context and reference files. The portion of code that the model cannot write is shrinking rapidly. Compared to eighteen months ago, the efficiency of implementing ideas and running research experiments has increased by twenty to fifty times. You can run multiple experimental ideas in parallel, and let the model monitor the runs and analyze the results. This has actually made my working hours longer and more intense; because the iteration cycle is so fast, there are constantly more ideas I want to test immediately. In the generative AI field, no one is coasting. If you are self-driven and passionate about the technology, the pace is incredibly demanding.

The absolute capability gap between Chinese and Western AI models is narrowing, and Chinese model developers have gotten incredibly good at building competitive models despite facing severe hardware and compute constraints. However, whether they can completely close the gap or surpass Western models remains an open question. The compute disadvantage has forced Chinese labs to become highly innovative in other areas, such as model distillation and multi-agent training systems. Model distillation is an open secret in the industry, but there is a major difference between brute-force distillation and smart, soft distillation.

Brute-force, or hard distillation, involves simply taking a massive batch of outputs generated by a model like Claude and training your own model directly on those tokens. Commercially, this is questionable, and intellectually, it is somewhat foolish because it shows that a team does not know what they want to build; they are merely copying someone else’s distribution to score higher on standard benchmarks. On the other hand, soft distillation involves using other models as assistants or evaluators within your own proprietary data generation pipeline. This is a fascinating scientific approach. It has effectively turned some Chinese labs into pioneers of multi-agent training systems, as they are integrating diverse model distributions into a single cohesive training ecosystem. ByteDance’s Doubao is quite distinctive; while it may not match the raw reasoning intelligence of Gemini or Claude, its real-time voice generation is arguably the best in the world. They have optimized heavily for their specific consumer user base, prioritizing low latency and engaging user experience over raw benchmark performance.

Before Claude 3.7, post-training was largely a process of patching up pre-trained models rather than a scalable paradigm of its own. The breakthrough came when the industry figured out how to scale reinforcement learning by utilizing clean, objective environments with clear feedback signals—specifically software engineering environments. The core development of Claude 3.7 took about four to five months. My work was focused on the algorithmic stability of large-scale RL and building the modular training environments. However, AI development at this scale is a massive, highly collaborative systems engineering effort. The era of individual heroism in AI is over; the technology is an unstoppable wave, and researchers are simply surfers trying to catch the crest. If one individual does not make a specific discovery, another team in a well-funded lab inevitably will a few weeks later.

Anthropic has a highly effective, top-down decision-making structure. Because the core technical leaders like Jared Kaplan and Sam McCandlish are also co-founders of the company, they possess immense technical credibility and corporate authority. When they receive a clear technical signal, they can instantly align the entire company to go all-in on a specific bet, such as prioritizing coding capabilities. This is very difficult for other companies to replicate. At OpenAI, internal political struggles disrupted this alignment, and at Google, the organization is historically much more bottom-up and distributed. Anthropic’s co-founding team has a deep level of mutual trust built from fighting battles together since their days at OpenAI, which keeps the organization incredibly reactive and focused.

I left Anthropic due to several factors. First, Anthropic grew very rapidly, more than doubling its headcount to nearly two thousand people while I was there. This rapid growth introduced significant cultural dilution and internal friction. I personally dislike cultures where people spend their days on Slack talking about grand, abstract principles rather than doing the hard, meticulous engineering work required to implement ideas. Second, I disagreed with the increasingly extreme and emotional anti-China political stance pushed by Anthropic’s leadership. Finally, Anthropic is highly focused on a narrow product set, meaning they do not invest in low-level infrastructure research or multimodal generation. I wanted to learn from a broader set of human minds and have the freedom to explore diverse research directions, and Google DeepMind offers an unparalleled environment for that scale of research freedom.

If your goal is to immediately force your specific idea into a consumer product, Google’s massive size and complex release processes can be highly frustrating. But if your goal is research freedom and accessing world-class engineering talent across a massive spectrum of technologies, there is no better place. Google’s technical reserves are unmatched. Additionally, OpenAI’s early success actually saved Google by waking the giant up. It forced Google to reorganize and streamline its decision-making, transforming pre-training into a highly disciplined, predictable engineering process that operates completely within Google’s comfort zone.”

This bodes well for the future of AI investment—scaling laws are alive and well, models are becoming intrinsically more intelligent, and so AI will continue to be able to solve a wider diversity of problems, and more complex ones. However, going forward, capex growth rates will moderate from the hyperfast expansion we’ve seen over the past few years:

Besides the law of large numbers, another key reason for moderating growth rates is that the four largest hyperscalers are now hitting their limits in terms of what they can afford without having to raise large amounts of debt or equity:

It’s likely that in the coming years, hyperscalers will continue to expand capex as operating cash-flow continues to grow, as raising large amounts of debt and equity to go free cash flow negative will likely impact their share prices, as we’re seeing with Meta for example.

Meanwhile, also neoclouds are getting stretched in terms of what they can spend on data centers:

Due to the huge surge in AI capex over the past three years, the strongest outperformers in the market have been the AI capex beneficiaries. However, in our view, the party isn’t over yet for semi investors as once the growth in AI data center capex slows down, it’s likely that a new S-curve will start to emerge in edge AI. One obvious application of edge AI is inserting intelligence into a smartphone.

Apple, which was still TSMC’s largest customer last year, knows that it has to get its AI strategy right. Missing the boat on edge AI—for example, with Google or Samsung launching a much better AI smartphone—could pose a big risk. In order to run a smart AI model on the edge, loads of DRAM are needed. This means that Apple will have to use its cash reserves to book new capacity at the three DRAM players. From the Wall Street Journal:

“Compounding the issue is Apple’s need for additional DRAM to support more AI features, including a rebooted Siri announced last week. And the company has long used NAND storage upgrades to boost profits, charging $100 to $200 for extra increments that cost it just a fraction of that. In the interview, Cook said Apple stands ready to use its cash reserves to boost memory supply. “We’re willing to use our balance sheet to help be a part of the solution,” he said. “Obviously, more capacity is needed.”

Mobile is still massive for the DRAM manufacturers—at 32% of revenues, Mobile & Client make the same contribution to Micron revenues as the cloud:

For investors wondering where to position in memory, our view continues to be that DRAM is much more attractive than NAND. DRAM is really the fast memory that enables AI, whereas NAND is basically storage where you store the chat history, or offload the KV cache as it grows too large.

In addition, DRAM is a highly consolidated three player market with more differentiation in HBM, whereas NAND is a more commoditized product with five significant manufacturers, increasing the risk for overcapacity at some stage. So, we expect DRAM to continue to enjoy the largest price increases:

Besides mobile, also physical AI (robots) is likely to become a large semis market in the 2030s. If, ten years from now, 100 million humanoid robots are sold per annum (vs a smartphone market which is currently at 1.25 billion units annually), this would translate into another semi TAM of $600 billion. For comparison, Nvidia will be doing $390 billion in revenues this year.

Yao Shunyu explains the current challenges still to develop humanoid robots:

“I have looked at some of the humanoid robots and their pricing; the hardware supply chain in China is incredibly mature, making these physical platforms surprisingly cheap. However, on the software side, robotics is still very much in its feature engineering era. Robots can be optimized to perform incredibly well in highly specific, deterministic environments, but they lack generalizability. If you train a robot to fold clothes or pour water, that capability does not naturally abstract and generalize to other tasks. Language models have crossed that critical threshold where scaling up the base model lifts all downstream capabilities simultaneously. Robotics has not yet reached its GPT-1 moment; we have not figured out how to truly scale the training of physical action models. Many robotics labs are now experimenting with using multimodal language models as a base, training vision-language-action models to bridge this gap. Visiting these labs is incredibly fun—they are far more dynamic than language model offices, with researchers constantly collecting physical data and controlling machines.”

XPeng’s founder goes deeper on the various challenges in robotics, although he estimates the company can start mass production of its humanoid robot ‘Iron’ next year already—these are the highlights from his interview on the Zhang Xiaojun podcast:

“What works for digital AI cannot simply be copied to physical AI. It is a completely different world. When digital AI companies try to define the physical world without ever having run a physical business, they build physical AI in a very narrow sense. The actual physical world involves complex human interactions, environmental variables, regulatory compliance, and commercial viability. You have to think far broader than just having a couple of strong features. That is why digital AI CEOs cannot easily explain the transition to physical AI; they are still discovering if they can even succeed.

XPeng’s robotics journey actually spans three distinct stages. The first stage was from 2018 to 2020. It was an independent team of about four or five companies in China exploring quadruped robotics. The second stage was from 2020 to 2023. Over those four years, we built three different milestones. We tried making robots using traditional robotics methods and even tried making robots the way we make cars, stitching various elements together with mixed success. The third stage began after 2023. When we saw the progress of foundation models in late 2022, our entire logic changed. Previously, we believed it was impossible to build a successful robot brain because the complexity of the cerebellum—maintaining physical balance and motion control—was too high.

Today, many companies claim they have developed a robot cerebellum because their robot can walk forward slowly at a monotonous pace. That is not a cerebellum; that is just a basic spine or brainstem maintaining balance. They are far from achieving true cerebellum functionality. Many of the robot demos you see from other companies today are still utilizing what we consider third- or fourth-generation technology stacks. They are just running basic tests. At XPeng, we have the patience and courage to invest for the long term. A quick demo means very little. It is just like back in 2017 when China had hundreds of autonomous driving startups showing off level 4 data; very few of those technologies ever translated into real, commercialized value.

Cars have a highly mature process spanning from planning and design through start of production. By the end of this year, we hope to transition our robots into an automotive-grade start of production process. By 2027, we expect high-level robots to enter their first true year of commercial mass production globally. At that point, traditional motion-controlled robots will begin to decline as advanced physical AI robots take over.

From the end of last year through the first half of this year, our robotics department alone hired nearly eighty master’s and doctoral graduates from top-tier institutions. They are incredibly expensive, but we are fully committed to supporting their long-term exploration. We believe that we must use super-smart people to solve super-difficult problems, rather than relying on rigid processes and predefined tools.

Starting a robotics business is vastly different from starting a car company. I believe starting a robotics company is twenty to one hundred times more difficult than building a car company. Even with XPeng’s established manufacturing and AI capabilities, our success last year only increased our overall probability of success by a small margin. Today, there are over two hundred robotics startups established in China, which is double the number of electric car startups we saw during the peak of the automotive boom. However, unlike cars, which are primarily categorized into passenger, commercial, or special vehicles, robots will have countless classifications, ranging from medical and freight transport to cargo inspection. Many of these specialized robots do not need to be humanoid. But on the path of universal bipedal humanoid robots, ninety-nine percent of companies will fail. The software complexity is simply too high, and there is no open-source platform capable of helping a robotics startup build high-fidelity physical AI software. We are focused on navigating these pitfalls ourselves because you only truly understand the engineering challenges once you step on the traps.

Traditional automotive motion control has been largely commoditized over the last century, with car companies purchasing motor controllers from Tier 1 suppliers. But to make a robot truly capable, its motion control must be far more integrated and responsive than a car’s. In a car, different control domains are isolated. This works for standard driving, but if a car has its left tires on snow and its right tires on grass, executing a forty-seven-degree turn with a person suddenly appearing in front of the vehicle is incredibly difficult. Managing four-wheel traction, movement balance, and latency under those conditions remains a challenge.

For a robot, the complexity is multiplied. A human has over two hundred joints, allowing for an infinite loop of potential movement combinations. Replicating this full-pose personification using AI-driven motion control rather than rule-based software is incredibly difficult. We want to give our robots the same physical instinct a human has when walking on snow, ice, or grass, sensing friction and adjusting balance instantly. This is why we develop eighty percent of our robot hardware in-house, including our actuator joints and hands, while cooperating with upstream Tier 2 suppliers to scale quality.

Once a robot achieves generalized software capabilities, its market adoption and production scaling will happen far faster than cars. Cars took a century to scale because we had to build global road infrastructure, establish traffic regulations, and manage highly complex manufacturing logistics. Robots can be deployed instantly into existing human environments. If the software is ready, physical AI will scale rapidly.

Over the next few decades, universal humanoid robots will become deeply integrated into human lives. While digital AI is limited to assisting in a few dozen white-collar roles, physical AI robots can address hundreds of physical roles, especially in aging societies. I believe the two most critical factors for the future of humanity are medical AI to help the elderly live longer, healthier lives, and physical AI robots to provide care and companionship. For many elderly individuals, a robot may eventually become their primary support system.

We chose this difficult path because we analyzed the fundamental flaws of other forms. For example, we previously developed quadruped robots—dogs and horses. But if you bring a robot dog that is over a meter tall into a standard home, it cannot function. It cannot turn around next to a bedside table without scratching the walls or damaging the bed. Unlike a real golden retriever whose tail and body are soft, a rigid robot dog will inevitably cause damage. If you make the robot dog smaller, its battery life becomes too short, and its utility drops to basic companionship.

If you build a massive, 1.8-meter bipedal robot covered in heavy metal armor, even its own designer will feel an intense sense of physical oppression standing next to it. You will naturally worry about it falling, overheating, exposing high voltage, or simply being dirty. If adults feel that way, how will children and the elderly react? How do you resolve those safety and legal regulations in a household setting? While industrial robots can utilize that design because they operate in controlled environments, household robots must be designed differently. They must be comfortable for humans to interact with. That is why our current generation of robots is designed to be around 1.69 to 1.70 meters tall—a height that is comfortable for both men and women. They are designed to wear clothes and can even have hair, but they must not have a realistic human face to avoid the uncanny valley effect and other complex sociological issues.”

Next, we will dive into a small company that’s an advanced silicon player in edge AI. We see the long-term risk-reward as very attractive in this name, and it reminds us somewhat of investing in Nvidia in 2015, when AI and data center GPUs were only at their very early beginnings. Eight years later, demand fully exploded. It’s likely that a similar scenario will play out when it comes to edge AI and robotics.

Read more

ASML’s tone has shifted from estimates where demand will be in the coming years to how much capacity they can provide (source: JP Morgan conference). Due to manufacturing optimizations, the company can probably ship 110 EUV tools in the coming years without building new clean rooms.

We can see that EUV shipments showed lackluster growth as from 2021. This was driven by two key factors—the industry digested the capacity which had been added during the covid semi boom while TSMC also moved to the 2nm node, which didn’t see an increase in litho intensity. Because TSMC didn’t dramatically increase the number of EUV exposures per wafer at 2nm compared to 3nm—as it moved from the FinFET architecture to gate-all-around (GAA)—the litho intensity on this new node flattened. Bears assumed that if new nodes won’t require steep increases in EUV anymore, ASML’s high growth era would be over.

However, as transistor scaling resumes post 2nm as we move into the Angstrom era with the A14 and A10 nodes, EUV usage node-on-node will again see substantial increases. The same is true for new DRAM nodes, and ASML actually expects the steepest growth in litho exposures node-on-node in DRAM:

At the same time, AI is massively DRAM intensive (especially in HBM), which is driving an unprecedented wave of capex growth in the industry:

While also TSMC will grow capex by 37% this year. We continue to think that the sell side is underestimating TSMC capex growth in the coming years. First, TSMC capex growth rates are still below growth rates during previous upcycles like 2019 and 2021:

Secondly, also TSMC’s capex-to-revenues ratio remains at cyclical lows:

We think that TSMC has a lot of gas left in the tank and that the risk-reward is strongly to the upside here for semicap names such as ASML.

Overall, we see a scenario with steep capex growth at both the DRAM players and TSMC playing out, while we also know that litho intensity at their coming nodes will continue to go up. Meanwhile, analysts continue to model fairly tepid revenue growth for ASML in the coming years, with much lower growth rates than during previous semi upcycles:

If we model in unit shipments from the JP Morgan conference, we already get to EUR 63 billion in revenues, 19% above consensus, giving a 27% revenue CAGR as from 2026:

Below is the model we originally built based on ASML’s capital markets guidance, although we added a very high scenario due to the current demand boom in AI. As a result, ASML will likely move towards EUR 64 of EPS much sooner than 2030, possibly already in 2028:

Bottleneck investing has become popular and ASML is an obvious bottleneck in the AI value chain—it is the monopolist provider of both EUV and immersion lithography equipment. We continue to be ASML bulls and hold this name.

Next, we’ll go through further findings in leading edge semis from this week, a number of stocks we’re purchasing at these levels, including new names, and our current top conviction list in semis.

Read more

The AI data center is splitting into two. On one side, you have Nvidia’s completely integrated kit—the extreme co-design that Jensen loves talking about. On the other side is the ASIC world where hyperscalers are designing custom solutions to get access to cheaper tokens, and to not be fully reliant on Jensen’s shop.

Marvell’s VP of Investor Relations is a semiconductor veteran and he tends to give much better insights on what’s going on at the company and in the industry, as opposed to CEO Matt Murphy who’s gotten a bit too salesman-like for our taste. All he does these days is talk about how pumped he is and how big the numbers are going to be in the coming years. So, the IR did a much better job detailing Marvell’s strategy in the AI data center at the Evercore conference, and we’ll go through some key insights below.

First, Marvell’s strategy is to become a bridge between the ASIC world and the Nvidia world in the AI data center:

“If you think about hyperscalers, they’ve really got 2 different infrastructures, and they don’t talk to each other. In the long run, that’s not how you want to build a data center. You want complete fungibility. That’s where Marvell comes in, with our networking IP and our custom IP. We have the bridge between both those worlds. Now, you can have hyperscalers being completely fungible in terms of how they design their network. So, we see that as really opening a much bigger TAM for the two companies combined [Nvidia and Marvell].”

A big theme in AI is scale-up—linking more XPUs together and make them behave like a single XPU—and Marvell’s goal is here to become an integrated shop which can offer all the various tech involved, whichever way the hyperscaler wants to go:

“Scale-up is almost a perfect example where a customer can work with us on the entire rack-scale infrastructure upfront. We can design all the chips for them, the entire signal path, give them different optimization paths, and this is a fairly unique ability. So, when you want to build a scale-up network, there’s 3 key components. There’s a compute engine, the XPU. There’s the interconnect, which is copper today going optical over time. And then the switch. We have all 3 pieces—we build XPUs, we’re investing in switches, and we have the leading photonics technology in the market.

On top of that, each hyperscaler is looking at multiple different alternatives, this is where our breadth of technology and investment lets us do multiple flavors. So we’re not making a bet on any one single thing. In optics, we have 3 different modulator technologies for doing scale-up optics. It’s MRM, Micro-Ring; MZM, Mach-Zehnder; as well as EAM, Electro-Absorption Modulator. We’re also investing in even more exotic technologies, whether it’s microVCSEL, or microLED. So, pretty much all options on the optical side are available. Even on copper, we can do co-packaged copper (CPC) if that’s what customers want initially as they go towards NPO and CPO.

Similarly, on the protocol side, on the switch side, we’re not restricted to offering one type of technology. We have a UALink switch coming out now, it’s 115T product. We have an ESUN (Ethernet Scale-Up Networking) product, which is based on a 100T platform. And of course, with NVIDIA, we have the NVLink platform.”

What should become a crown jewel in interconnect is recently acquired Celestial AI, which developed a photonics fabric that can connect hundreds of XPUs across racks and make them behave like a single XPU:

Celestial AI has a big win with a major hyperscaler (Amazon) and its photonics solution is on track to ramp in ‘27 and ‘28, with already $1 billion of guided revenues for calendar ‘28.

Deepening its portfolio of advanced optical tech, Marvell also acquired Polariton. This is a spinoff from ETH Zurich that has developed a unique solution to increase modulator speeds with 10x.

The bear thesis on Marvell over the past years was that their stronghold in optical PAM DSPs—where Marvell dominates with a 70% market share—would get disrupted by Broadcom. However, we’ll make two key points here. First, this is a high growth market and so even with share losses this would remain a high growth business for Marvell. Secondly, hyperscalers want multiple suppliers and so we were always skeptical on the thesis that hyperscalers would shift even more of their spend towards Broadcom, which is already the gorilla in datacenter networking.

On the other hand, we see hyperscalers keen to work with a variety of niche silicon suppliers, including Astera, Credo, etc. So, our thesis remains that hyperscalers will continue to spread their budget among this wider group of networking silicon providers, as opposed to consolidating their spend with Broadcom. Marvell’s IR mentioned how their interconnect business remains on fire:

“Our interconnect business is growing at 70% plus for this year. Within that, there’s really 2 big underlying drivers. There’s a scale-out business that’s growing even faster, that’s our PAM DSP and TIA driver business. And then you’ve got scale across, which is just starting and is going to get a lot bigger in the outer years. So, the short to medium term from a revenue perspective looks very strong. And then what’s even more exciting is all the engagements we are seeing a little bit further out in time.

On competition in DSP, the basis of competition in these markets is always “were you first to market?” That’s been the case in every single generation of PAM4, and it’s no different at 1.6T. The largest part of our interconnect business is our PAM DSPs, and next year, 1.6T is a huge part of it. The reality is I don’t think we see anything significantly different. We’ve had a leading position, and I don’t really see that changing going forward.”

Marvell is supplying the DSP, the driver, and the TIA:

And with silicon photonics the company can even sell an extra die (a SiPho PIC, or Photonic Integrated Circuit):

The other bear thesis on Marvell was that they didn’t know what they were doing in custom ASICs and would lose those programs to Broadcom. However, also this bear thesis took a massive dent over the last months as Anthropic is currently training the most advanced LLMs on Trainium clusters. At the same time, Amazon is massively scaling up its AI capacity and is doing this on a mix of Trainium and Nvidia infrastructure. The key platform in AWS’ AI strategy is Bedrock and that is largely running on Trainium. Hyperscalers are extremely good at writing intermediary software layers to run workloads over a variety of underlying hardware, something which will only become easier with Claude Code.

Thus, all data points are pointing to Trainium being a massive success, and Marvell will remain a key partner in this program going forward. JP Morgan notes that Marvell is working on the design for the 2nm Trainium 4 chip, as well as a large number of other ASIC programs including the Microsoft Maia XPU:

“In custom AI ASICs, the next-generation Trainium 3 XPU ASIC program is starting its initial shipment ramp this quarter, with volumes increasing in the second half of the year. We believe the team [Marvell] is deep in the design phase of the next-generation Trainium 4, 2nm program. Their next Tier-1 XPU customer (Microsoft, Maia 3nm XPU program) is progressing well, with plans for high-volume manufacturing next year. Relative to the Microsoft demand forecast, we believe Marvell is still being very conservative (FY28) and taking a significant haircut to the actual forecasted numbers. The team has 10+ XPU attach programs ramping next year (DPU, CXL, networking). Overall, we are impressed with the strong multi-year revenue outlook and the diversity of customer program ramps, and we see a solid setup for CY26/ CY27.”

Evercore’s background checks point to Marvell being a key partner for these programs:

“Over the last 6 months, we talked to about a couple of dozen sources amongst the hyperscalers. And very consistently across the group, what we heard is that Marvell is being viewed more as a more strategic partner supplier. And that’s because the broad portfolio of IP, not just on the XPU side, but also on the networking side.”

Subscribe now

Marvell’s IR explains their role in XPU design, versus where competitors in Taiwan such as Alchip play:

“The reason why customers are coming to us is because there’s a lot of our interconnect technology when you build large complex XPUs. These XPUs are no longer monolithic single-chip devices. These are multiple compute dies and HBM stacks. So you need high-speed SerDes, you need die-to-die interfaces, you need custom HBM interfaces, you need much more optimized custom SRAM for much higher packing density, you need advanced packaging, and those are all things we do for our merchant business. So the reason we’re in the custom business is because of the expertise we’ve established from our merchant business.

Now when you build a chip, there’s an IP portion, which is more considered kind of front-end design, but right at the end of the process, you have to do layout, which is physical design. So some of the design service companies have an important role in the ecosystem, but they don’t have IP. And the reason they don’t have IP is because they don’t have a product business. That’s the real distinction. If we are engaged in a project, it’s because our networking IP is what the customer is looking to access and build into the XPU, versus when they’re partnering up with somebody in the design services side, it’s more of a relationship where you need someone to do the last part of the process, which is physical design.”

One example of specialized IP is Marvell’s SRAM with higher packing density, which enables 17x the bandwidth versus off-the-shelf IP according to the company. Marvell illustrates how this will be useful for both 2nm accelerators and CPUs:

And then also XPU attach is a big opportunity for Marvell. XPU attach is basically companion chips that can take background tasks over from the XPU, allowing the XPU to focus on key AI workloads. This is Marvell’s IR on this market:

“Think about XPU attach as an offload device. You’re trying to maximize your central compute device for its core function. This is no different than years back, where you started creating NICs to minimize I/O of the server CPU. Same idea essentially. CXL and NICs are the two large examples, but there’s a couple more design wins going into production. One would be storage accelerators. Whether it’s SSDs or HDDs, instead of having the core CPU or XPU address them, you’d rather do it on a dedicated device. So that’s a storage accelerator.

And another one is the security offload device. This is a unique product for Marvell, and comes from the Cavium acquisition. We’re one of the only companies building dedicated security accelerators. The idea is that I want to make sure my data when I put it into a cloud network is encrypted. You can certainly do encryption/decryption on a standard XPU or a standard CPU, but why would you? You’d much rather do it on a dedicated device. So that’s another example, which is again going to be a fairly large opportunity for us.”

Millions of XPUs are sold annually, and so if you can cross-sell $1,000-$2,000 worth of silicon that goes alongside these XPUs, this is obviously a big opportunity. Marvell is guiding for $3 billion of revenues in FY28 to come from XPU attach. Both CXL and custom NICs are $1 billion opportunities each.

And this opportunity continues to increase with agentic AI—if you need more CPUs, you need more NICs. And as agentic AI creates more KV cache data, this means also CXL will be needed. CXL is basically an interconnect that lets XPUs treat external DRAM as if it were directly attached to the processor. This way, excess KV cache can be moved from HBM to CXL-attached DRAM.

Share

Overall, we like Marvell’s strategy. The company is clearly building out a portfolio of some of the most cutting edge tech in both AI data center networking and custom silicon. Recent acquisitions like Celestial AI and Polariton are good examples of this strategy and seem once again like very smart strategic moves by Matt Murphy. So, we liked the risk-reward in this name a lot over the last twelve months, as the shares were trading in the low twenties on a forward PE basis:

However, after Jensen decided to pump the stock in Taiwan, we’ve now moved to a 70x forward PE valuation. For the bulls, we actually don’t think the shares are massively overvalued now—it’s likely that numbers are still too low given the emergence of agentic AI, and so we can move towards $10 of EPS already in calendar 2028. This gives a current valuation of around 30x 2028 numbers.

However, this valuation now warrants absolutely flawless execution. At the same time, we suspect that a lot of the speculative and fast retail money has now moved into the stock—the kind of crowd that buys based on a meme or a news headline. In our view, these two factors make the risk-reward now unattractive, with likely a substantial correction again at some stage. After shares rose 30-35% post Jensen’s comments, we decided to liquidate our position. We do love Jensen, somehow this guy always figures out a way to make us some nice profits.

Server CPUs Are A Hot Market Again

At Computex, Nvidia illustrated how the CPU has become important again in agentic AI workloads:

However, it’s also clear that the accelerator (GPU/XPU) will remain dominant as this is where all the reasoning happens. On the other hand, when the AI has to make a database call, do a web search, or execute a Python script, this happens on the CPU. So, the GPU writes the Python script, and then the CPU executes it.

Jensen highlighted at Computex how CPU performance becomes crucial in agentic workloads, from a transcript created by Claude:

“Huang argued that all CPUs until now were created for people — humans who live in a world counted by seconds, who rent CPU cores in the cloud at hourly rates. Agents are fundamentally different. They are impatient. They live in a world counted in nanoseconds. When an agent uses a tool or accesses a database, the response must come back as fast as possible. Every moment of waiting keeps the agent from proceeding to the next step. And because CPUs sit in the critical path next to extremely expensive GPU infrastructure that generates token revenue, these CPUs must be both high-performance and highly energy-efficient.

A narrated video elaborated on the Vera CPU’s technical architecture. The NVIDIA Olympus core at the heart of Vera is built for modern data center workloads including branch-heavy Python runtimes, tool calls, and sandbox code execution. The video described Vera achieving 40 percent lower peak memory latency versus x86. Memory-coherent NVLink chip-to-chip connects GPUs directly to the CPU and can also scale Vera up to multiple sockets. Vera delivers 1.8 times the agentic sandbox performance of x86 CPUs.

Returning to the stage, Huang presented benchmark results. He showed SQL running three times faster on Vera—calling it extraordinary since SQL is among the most difficult workloads to accelerate. He also presented real-time stream processing results for the New York Stock Exchange, where Vera CPU runs six times faster, crediting the bandwidth improvements, single-threaded instruction execution, and internal and external bandwidth improvements of the architecture.”

UBS notes that Nvidia will be doing more CPU revenues than AMD this year, which also includes standalone CPU sales for Nvidia:

“Maybe the most controversial item was new disclosure that “standalone” Vera CPU will be ~$20B in revenue this year and opens up a new $200B TAM. On the surface, this is a much larger number this year than we would have expected and is even bigger than AMD which we model to be ~$16B this year and has >50% share in the cloud segment. In speaking with the company, there are some clear use cases that are standalone chips (e.g. CPUs attached to non-NVDA accelerators, for example, META as part of its recently signed Vera CPU deal) but this also includes “standalone” systems that only have a CPU with no GPU. It is hard to say, but in these configurations, the standalone CPU chip might be ~30% of the total value with DRAM being another large component. So NVDA may only be doing ~$5-7B of CPU chip revenue this year - which would imply ~1-1.5MM units - which makes sense to us. Regardless, NVDA’s forward thinking on supply should especially come to bear in this agentic world where newcomers like ARM and even QCOM are competing for capacity and we read this as negative for the aspirations of these other competitors.”

Subscribe now

Similarly, Qualcomm—the leader in power efficient smartphone CPUs—is now looking to deploy their skills and take share in the data center CPU market:

“The second area where we’ll be entering the data center is the CPU. As you might know, we have a custom CPU that we deploy in handsets, where we are the performance leader. We also deployed that same CPU in PC. And so if you compare us to the x86 players, Intel or AMD, we think we have a very significant performance advantage. And we’re going to bring all of that to bear in a data center CPU solution. CPU use is changing in the data center, and it’s a very large market.”

Even ARM is looking to compete with their own customers and has now launched its ‘AGI CPU’:

“As AI is moving from human-based queries to continuous agent-driven workloads, this shift is expanding the role of the CPU. These Agentic workloads require CPUs to coordinate tasks, move data, manage memory, enforce security and orchestrate workaround accelerators. As Agentic AI scales, data centers will require more than 4x today’s CPU capacity, creating a data center CPU market opportunity of more than $100 billion by 2030. The Arm AGI CPU, which we launched at our Arm Everywhere event last quarter, is purpose-built for Agentic AI.

Our first production silicon product for the data center will deliver more than 2x the performance per rack compared with x86 platforms with the potential to reduce AI data center capital expenditure by up to $10 billion per gigawatt. Meta is our lead partner and co-developer, and is working with us on a multi-generation road map to support personal super intelligence for more than 3 billion users.

Customer response to the Arm AGI CPU has been very strong. We now have more than $2 billion of customer demand across fiscal 2027 and fiscal 2028. This is more than double what we stated at launch. We are on track towards our forecast of $15 billion as stated at our Arm Everywhere event. And soon, the data center will be Arm’s largest business.

For example, SAP will move their core database and business application workloads to Arm, starting with AWS Graviton and expanding to the Arm AGI CPU. This represents a significant strategic shift. Cloudflare will deploy Arm across its global network to support traffic management, security and AI inference closer to users. We have also secured design wins with key network infrastructure providers, including F5 and SK Telecom.

AI infrastructure needs CPUs and accelerators working together efficiently at scale. NVIDIA, Amazon and Google are already using Arm-based CPUs as head nodes along the accelerator-based systems. Cerebras, OpenAI, Rebellions and Positron are doing the same with the Arm AGI CPU. This momentum builds on our existing scale in the cloud. That scale is increasingly driven by Arm Neoverse CSS and Arm-based compute, which now represents about 50% market share with top hyperscalers.”

Historically, Intel and AMD had the x86 moat which meant that all server and PC software was written to run on x86 cores. However, due to smart container-based software such as Kubernetes, the large clouds can now deploy apps inside containers which abstract away the hardware underneath. This means that apps can run on both x86 and ARM-based CPUs, which is allowing hyperscalers to move workloads to their custom ARM-based silicon.

We’re only at the start of this trend as only Amazon has scaled its Graviton CPUs so far, but also Microsoft and Google are looking to replicate this strategy. Where x86 still has a stronghold is in the enterprise server market (i.e. on-premise data centers), however, cloud is the high growth market and over the long term, we suspect that most of the public cloud will move towards ARM cores.

Thus, cloud CPUs will become an extremely competitive market. Nvidia has already become massive in data center CPUs, in the blink of an eye. Another giant with Qualcomm is entering as well, which has strong skillsets in power efficient CPU design. Even Arm has cobbled together a number of interesting deals such as with Meta and Cloudflare. At the same time, hyperscalers such as Amazon, Microsoft, and Google will increasingly move cloud workloads to custom silicon.

Share

Despite increasing competition, we still see two key positives for AMD. While all hyperscalers are mainly looking to deploy a Nvidia + ASIC strategy in accelerators, AMD as a third player in the market can still make substantial money. Its GPUs with higher HBM capacity versus Nvidia are an advantage in inference, so we continue to think that AMD will be able to grab a smaller share in the GPU market, which will still be lucrative. Deals with OpenAI and Meta illustrate how the company has indeed secured share.

At the same time, momentum in the server CPU market will be tremendously strong in the coming years, while newcomers such as Qualcomm and Arm will only gradually gain share. Microsoft and Google are also still only at the start of their CPU ASIC journey. Thus, we see both AMD’s CPU and GPU businesses as booming in the coming years.

However, at 60x forward EPS, and 29x 2028 EPS, this strong story is now well reflected in valuation. So, we don’t see the risk-reward as being attractive at this stage. We could be interested again later in the year if there is a strong correction in AMD.

A negative for AMD is that capacity is extremely tight—TSMC isn’t able to meet all demand and this will also limit AMD to secure additional capacity for the boom in CPUs. This is AMD’s IR at the BoA conference:

“Supply is tight. 3-nanometer is tight. There’s a lot of other areas that are tight. I think we’re very well positioned with the relationships Lisa has personally with the executives in that space to get maybe more than our fair share of the incremental. But it is tight.”

It’s just something to keep in mind as it will limit AMD’s potential to surprise on the upside in the coming years.

Next, we’ll go through a variety of names that we see as winners from agentic AI—we’re currently halfway through the analysis.

Read more