Anthropic IPO Analysis: Compute Commitments and Margin Volatility Outweigh Valuation
Summary
Vincent analyzes Anthropic's IPO at a ~$1.3 trillion market cap, noting its rapid revenue growth and 55% enterprise LLM API market share. He highlights the company's diversified chip strategy and positive operating margins since Q2 2026. Despite strong cash reserves and gross margins, Vincent warns of volatility due to long-term compute contracts versus consumption revenue. He identifies massive compute commitments and sovereign AI trends as key risks. Consequently, he declines to buy the IPO, waiting for hype to subside.
Key takeaways
- 1 The 'Anthropic Direct' model involves leasing empty data center shells and supplying own chips, a strategy to secure scarce power infrastructure early while retaining flexibility to choose chip suppliers later, reducing reliance on short-term hyperscaler contracts.
Unidentified speaker (speaker_0) · 26:40
Episode sections
- 0:06 – 3:32
Anthropic IPO Overview and Growth Metrics
John Gillen introduces the episode, noting Anthropic's IPO at a ~$1.3 trillion market cap, making it the 12th largest company globally, with up to $215 billion in cash post-IPO and positive operating margins. Vincent begins his analysis by highlighting Anthropic's rapid revenue growth, citing a 12.3x increase over 3.5 years and a 47% quarter-over-quarter growth in the last quarter.
- 3:32 – 12:23
Revenue Efficiency and Enterprise Market Share
Vincent analyzes revenue per dollar of compute, noting a dip from $2.10 to $1.72 due to short-term contracts with providers like SpaceX, but argues this remains positive ROI. He highlights Anthropic's 55% share of enterprise LLM API spending versus OpenAI's 31%, attributing this to a B2B focus that creates sticky, high-net-dollar-retention customers.
- 12:23 – 17:46
Margin Analysis and Profitability
The discussion covers Anthropic's gross margins holding at 86% in Q3 2026 despite commoditization pressures, driven by cheaper models like Haiku 5.1. Operating margins turned positive in Q2 2026, but Vincent warns they will remain volatile due to the mismatch between long-term compute contracts and consumption-based revenue.
- 17:46 – 24:06
Compute Strategy and Chip Diversification
Vincent details Anthropic's diversified chip strategy, using AWS Trainium (50%), Nvidia, Google TPUs, and AMD. He notes that running Claude on AMD chips took only three days, providing leverage in negotiations. He contrasts this with competitors relying 100% on Nvidia, highlighting Anthropic's ability to switch suppliers quickly.
- 24:06 – 26:40
Balance Sheet and Compute Commitments
Anthropic holds $200-$215 billion in cash, generating $1.2 billion in net interest income in Q3. They have secured 50% of their $451.6 billion in compute commitments through 2030, with an additional $450 billion in commitments lined up for 2031-2045, making compute the primary business driver and risk.
- 26:40 – 28:48
Anthropic Direct and Cost Reduction
Vincent explains the 'Anthropic Direct' strategy, where the company leases empty data center shells and supplies its own chips, reducing reliance on short-term contracts with hyperscalers. This approach aims to secure scarce power infrastructure early while maintaining flexibility in chip procurement to lower costs and improve margins.
- 28:48 – 32:21
Growth Levers and Valuation
Growth drivers include automation of daily operations, agentic tools via partners like Salesforce, and AI in R&D budgets. At the $74 IPO price, Anthropic trades at 8.2x sales for 2027 estimates, which Vincent considers fair relative to private valuations of OpenAI and SpaceX, though he remains cautious on the stock.
- 32:21 – 37:54
Reasons for Skipping the IPO
Vincent declines to buy the IPO due to four risks: massive compute commitment uncertainty, sovereign AI trends where companies build internal models to protect IP, intense competition leading to commoditization, and the need to maintain frontier status via heavy compute spending. He waits for hype to subside and the price to drop.
Assets discussed
| Asset | Kind | Stance | Context |
|---|---|---|---|
| Anthropic | private company | neutral | First, anthropic. in comparison to for instance a SpaceX which relies 100% on NVIDIA or a meta which also relies I think 100% on NVIDIA but I'm on the cheap use but I'm not not fully sure on that or an OpenAI which also has a more than I think 50 or 60% in in OpenAI is |
| SpaceX | private company | neutral | First, anthropic. in comparison to for instance a SpaceX which relies 100% on NVIDIA or a meta which also relies I think 100% on NVIDIA but I'm on the cheap use but I'm not not fully sure on that or an OpenAI which also has a more than I think 50 or 60% in in OpenAI is |
| OpenAI | private company | neutral | First, anthropic. in comparison to for instance a SpaceX which relies 100% on NVIDIA or a meta which also relies I think 100% on NVIDIA but I'm on the cheap use but I'm not not fully sure on that or an OpenAI which also has a more than I think 50 or 60% in in OpenAI is |
| MicrosoftMSFT | equity | neutral | So think about today you have, for instance, Microsoft Copilot, which where you can select, you use the OpenAI models as a user, as a copilot user, you can choose which model you use, right? |
“Stance” is how the speakers talked about the asset, not a recommendation.
Claims worth verifying
Anthropic's IPO valuation is around $1.3 trillion, making it the 12th largest company in the world.
Unidentified speaker (speaker_1) ·0:26
How to verify: Check SEC S-1 filing for Anthropic's IPO price and shares outstanding to calculate market cap; compare to S&P 500 or global market cap rankings. For recent events, verify 2026-08-20 through 2026-11-20. Use another period only when the speaker explicitly names it.
Primary sourceAnthropic will have up to $215 billion of cash after the IPO.
Unidentified speaker (speaker_1) ·0:27
How to verify: Review SEC S-1 filing for pro forma cash balances post-IPO. For recent events, verify 2026-08-20 through 2026-11-20. Use another period only when the speaker explicitly names it.
Primary sourceAnthropic captured 55% of enterprise LLM API spending, versus OpenAI's 31% and Google's 9%.
Unidentified speaker (speaker_0) ·7:51
How to verify: Verify against Anthropic's S-1 filing or cited market research reports on LLM API market share. For recent events, verify 2026-08-20 through 2026-11-20. Use another period only when the speaker explicitly names it.
Primary sourceAnthropic's gross margin was 86% in Q3 2026.
Unidentified speaker (speaker_0) ·12:43
How to verify: Check SEC S-1 filing for Q3 2026 financial statements. For recent events, verify 2026-08-20 through 2026-11-20. Use another period only when the speaker explicitly names it.
Primary sourceAnthropic's operating margin turned positive in Q2 2026.
Unidentified speaker (speaker_0) ·14:49
How to verify: Check SEC S-1 filing for Q2 2026 financial statements. For recent events, verify 2026-08-20 through 2026-11-20. Use another period only when the speaker explicitly names it.
Primary sourceAnthropic spends about 50% of its compute on AWS Trainium chips.
Unidentified speaker (speaker_0) ·20:21
How to verify: Review SEC S-1 filing for procurement details or footnote disclosures on chip suppliers. For recent events, verify 2026-08-20 through 2026-11-20. Use another period only when the speaker explicitly names it.
Primary sourceAnthropic can run Claude on AMD chips in three days with no human intervention.
Unidentified speaker (speaker_0) ·22:04
How to verify: Check SEC S-1 filing footnotes regarding technical capabilities and chip compatibility. For recent events, verify 2026-08-20 through 2026-11-20. Use another period only when the speaker explicitly names it.
Primary sourceAnthropic has $200-$215 billion in cash and $451.6 billion in compute commitments through 2030.
Unidentified speaker (speaker_0) ·24:27
How to verify: Review SEC S-1 filing for cash balances and long-term lease/commitment disclosures. For recent events, verify 2026-08-20 through 2026-11-20. Use another period only when the speaker explicitly names it.
Primary sourceGoldman Sachs forecasts Anthropic will reach $600-$700 billion ARR by 2030.
Unidentified speaker (speaker_0) ·29:07
How to verify: Search for Goldman Sachs research reports or press releases from 2026-2026 regarding Anthropic forecasts. For recent events, verify 2026-08-20 through 2026-11-20. Use another period only when the speaker explicitly names it.
Primary sourceAnthropic trades at 8.2x price-to-sales for 2027 estimated sales at the $74 IPO price.
Unidentified speaker (speaker_0) ·31:27
How to verify: Calculate using IPO price ($74) and shares outstanding from S-1, divided by 2027 revenue estimates cited in the filing. For recent events, verify 2026-08-20 through 2026-11-20. Use another period only when the speaker explicitly names it.
Primary sourceDo your own research
- Review Anthropic's S-1 filing for detailed risk factors regarding compute supply chain dependencies and customer concentration.
- Verify Q3 2026 financial results against SEC filings to confirm gross margin stability and operating margin positivity.
- Analyze public data on enterprise LLM API spending shares to validate the 55% market share claim versus OpenAI and Google.
- Check Anthropic's disclosed compute procurement contracts and power infrastructure leases for 'Anthropic Direct' commitments.
- Research sovereign AI trends and corporate internal model development strategies to assess the risk of reduced third-party API usage.
- Compare Anthropic's 8.2x sales multiple against historical SaaS and AI company valuations during similar growth phases.
- Investigate the specific terms of short-term vs. long-term compute contracts to model revenue volatility scenarios.
- Monitor Anthropic's model release cadence and R&D spend to ensure continuous innovation matches the 80% revenue concentration in recent models.
Full transcriptexpand
0:00Unidentified speaker (speaker_0): There will be times where I think you get more attractive prices, and that's what I'm waiting for, and then we'll see whether I'll buy into it.
0:06Unidentified speaker (speaker_1): What's up everyone? I'm John Gillen. Welcome to Milk Road Stocks, the daily market show that went through Anthropic's entire S1 with a fine-tooth comb and a magnifying glass, and somehow forgot to read it. I'm your I'm your host, John Gillen. Today is Friday, November 20th. We recorded this episode on November 17th. Anthropic is hitting the market. They just IPO'd at around 1.3 trillion. of market cap, which makes it around the 12th largest company in the world. It just turned a positive operating margin and it will have up to $215 billion of cash after the IPO. And yet, after going through over 20 different charts, Vincent is passing for now. Today, Vincent, one of our leading analysts, is on the show to walk through everything in Anthropic's S1. He goes through this in detail, even deep into the footnotes on the S1, which I thought was really impressive. He's gonna break down all of his thoughts on this, and then he's gonna conclude with why he is sitting out the IPO and what he is waiting for instead. And speaking of investing, imagine having an AI that analyzes opportunities like a hedge fund analyst. That's why we built Portfolio OS. It connects your
1:07portfolio and investing style to the AI you already use, giving you institutional grade research tailored to you. Right now, lifetime access is 60% off, and you'll also get three months of Milk Road Pro for free. So check out the link in the description to get in on that. Without further ado, enjoy the episode. We'll see you in Milk Road Pro and thanks so much for being here.
1:26Unidentified speaker (speaker_0): Welcome back. It's the week for Anthropic IPO. And obviously that's what we're going to talk about, covering everything from the growth to the their margins to I have 20 plus. charts with me today. So uh yeah covering this from all different angles and then obviously in the end I'm also giving my view on whether I will buy into the IPO or not and what the reasons are for that. Let's let's jump directly into it. First let's focus on on what Anthropic did over the last couple of um quarters and years in terms of growth. It's pretty ridiculous if you think of it's been a startup um a couple of years ago and it's IPO ing it their IPO is valued at one point three three trillion between one point two to one point four trillion is is is rumored. So uh at the midpoint that will put them at the number twelve spot of being the largest company in in the world, right? So that's pretty remarkable. and tells you just how
2:29fast the scale of this company was. Obviously this is also reflected in their revenue growth. And I think this is I think one of the most important uh data points that investors will look at um and already looked at in the past couple of years and will look at once once they're public is is their growth and will their growth slow down or not? Um I I think both sides have fair arguments that they could slow down. In the end I'll have a full slide on how they could grow, um what the free core revenue drivers are and what the risks to that is. But just looking at the raw data means the fastest revenue ramp ever recorded at this scale, it's a 12.3x in just three point five years after Cloud's commercial launch, um, at a basis of one point four billion roundabout, which is just um ridiculous, right? And even uh in in in the last quarter they grew forty seven percent
3:32quarter over quarter. Um obviously there will be core moving forward. This I think is the even more interesting chart on revenue side. What this is showing is anthropic revenue per one US dollar of compute that um they're spending. And this is showing you that, or basically how to read this chart is every number above one is positive for um Anthropic. And this is showing you that, or this is proving that there is our ROI on their um CapEx investments. Now from Q2 to Q3, there was a slight dip from around $2.1 dollars to $1.72. Um, and that was because of the uh short-term compute contracts that they have, right? Uh, among others with SpaceX. Obviously, SpaceX is benefiting from that. Um, but I expect that. to to to to or this I expect this chart to remain volatile over the next couple of quarters because as you will see later, compute is the number one driver for entropic business. So they will not slow down in demanding compute. And they they are not
4:47as fast themselves by bringing compute online so they need to rely on short-term compute track contracts with a provider like SpaceX, and obviously that is really expensive, right? And that's why this chart remains volatile. But everything in the range, let's say, between 150 to and and and more is, I think, really positive because again, it's proving that there is ROI on their uh investments. This is also key to understand. About 12,000 customers generated at least 100,000 US dollars each over the last um 12 months. Um I think you can read this two ways. On the one hand, from the more bearish side, you could say okay, 47% um of their uh total. revenue is dependent on their biggest customers, which I guess is is is fair. It's just the 80-20 share that most companies have, right? But of course, if I mean there are many reasons why companies that use Cloud would switch
5:56providers or build their own models, um use open source, for instance, right? So obviously they have a certain dependency on a smaller share of um customers. That's kind of the bearish read. The more bullish side, which I would subscribe here to, is Anthropic has this clear focus or had this clear focus from the very beginning to focus on more the B2B side versus the B2C side, right? OpenAI, for instance, was always more focused on selling their products to us, basically, right? The subscriptions that we're all using today. Whereas Cloud was always more focused on the enterprise solutions. And enterprise customers are, of course, much more sticky. They're not as likely to switch suppliers because for them it's once you have the tool and and and a product. implemented in in all the business processes it's super annoying to get it out it's also super expensive to get it out out and and and as long as that product is um reliable to a certain extent
7:03reliable and is promising on on on on on the services then and and delivers a certain quality then it they will not switch and this all leads to a higher net dollar retention rate, right? And that's how I would read that chart. It's not that they have a concentrated customer base. Yes, they do, but those customers are increasing what they're spending um every year, right? And that's I think core core to understand here, because of their B2B focus. Another chart proving that therefore c that that Anthropics focus is on kind of the B2B side. They're the number one LLM in the enterprise um context. I think the numbers are pretty staggering here. Anthropic has a 55% share of enterprise LLM API spent, versus OpenEI having 31%, Google 9%, and the rest in others. Now this long term will probably switch around a lot as companies are trying to reduce costs by moving to open source. So
8:11to me, this is just a snapshot of how it looks today. But I think also seeing the year-over-year growth in the share that Onthropic has from 40% in 2025 to 55% in 2026 is showing you that. most of the companies out there that are heavy users of AI are relying on Cloud. I mean, look at all the big consultancy companies, right? It's publicly known. BCG is heavily relying on Claude, the investment banks, Eli Lilly, so all those massive AI adopters, AI-first AI first companies are um focused. focusing and using plots and and the tools and products from Tropic.
9:00Unidentified speaker (speaker_1): What if the AI you already use could get an upgrade and turn into a personal investment analyst that knows your portfolio better than you do? Introducing Portfolio OS by Milk Road. Connect your investments, bring everything you own into one place. and immediately get a risk analysis on your portfolio. Then build your investor DNA. Teach your AI how you invest and how much risk you're taking. Alpha Lab gives your AI the skills to find opportunities, research companies, challenge your assumptions, and evaluate what they're worth. And because it knows what you already own, Portfolio OS can help spot risks or opportunities you might otherwise miss. From your first investment idea to your exit plan, Portfolio OS helps you follow a disciplined research process, just like the pros do, but without the big budgets and big teams behind it. This is the future of investing. For Black Friday, get lifetime access to Portfolio OS for just $1.99 plus three months of MilkRoad Milk Road Pro free. Portfolio OS by Milk Road, your investing process upgraded.
10:07Unidentified speaker (speaker_0): From a revenue perspective, I think it's key to understand, and it's my view, that Anthropic was the number one winner of the Chansen fallout this week at NVIDIA. Right, you probably all know Chanson Chansen is out at NVIDIA for many reasons, among others, selling. chips to China through companies that are, for instance, Megaspeed, the Singaporean-based companies that is buying the NVIDIA chips out of Singapore and then providing the capacity to Chinese customers and even putting some chips into customers. Now, obviously, we've already seen that not only Chansen is out, but the export controls from the US
10:50Unidentified speaker (UNKNOWN): to
10:51Unidentified speaker (speaker_0): to China and other countries increase massively this week. And what this likely means is that the entire open source development or or trade, you could also say, is slowing down. With Chansen out, I mean he was the number one preacher of open source. It's understandable because NVIDIA for NVIDIA the entire open source kind of movement was basically bringing up a new customer next to the hyperscalers and Frontier AI. But now with that, I wouldn't say gone, but at least slowed down, Anthropic can benefit not only from a market share perspective, but also from a pricing powers perspective. And um that that I think is key to to understand. You know, obviously that is the situation how it looks right now, right? Just 20 or 30 minutes before I jumped on this podcast, Jensen posted on X that he's very open to come back as CO. And that could change. Um that view could change again, of course. But as long as Jensen Chanson is out of Nvidia, I would read that as bullish for
12:03Frontier Eye and bearish for the entire open source um kind of movement, especially the Chinese side of things. Now I think we covered the revenue side pretty well, and it's it's it's straightforward that the company grew massively over the last couple of years, and that there is ROI Eye on the investments. Now the question is: are they already profitable? How are their margins looking? like? And um what are they earning, right? So first from a from a gross margin perspective, right? So basically price and volume, it seems pretty attractive because even on a higher revenue base, much higher revenue base, if you compare first half twenty twenty six to Q free twenty 2026, twenty six, they could hold their margin in the mid eighties, 80s, eighty 86% six percent in Q Q3 free twenty 2026. twenty. Now, the deeper read of this chart, I think, is that they can hold their gross margin amid the entire commoditization argument of LLMs, right? This commoditization in this context basically means, and I think Alex Carp from OpenAI is is is
13:12is is someone that is very um loud about. this, is saying that there is no value in LLMs because the price will trend to zero over time. And we have seen that, right? The cost of LLMs per token decreased massively. But what we're also seeing is that the volumes are holding up and obviously gross margin a function of price and volume is um is is is is is holding up for the tropic. And that's I think the core read from this chart that at least up until now, commoditization is not in is impacting them, of course, but they can still make attractive gross margins. Um and and and and how they're doing that is basically offering cheaper models. Um like Haiku 5.1, for instance, uh, or 5.5 is the current model, I think. Um, that's that's the cheapest one. Um, so uh that is much cheaper per token, but still has high quality, right? And over time, they will keep providing new models um that are at the frontier that will remain super expensive, and then the models that are now at the frontier will become the kind of cheaper. um alternatives, right?
14:29And and and and and that's what we're seeing from that. Second, from a margin perspective, is the operating margin. And since 2026, Q2 for the first time that is positive. And obviously, kind of from a long-term sustainability perspective of this business, this is this where this is very vital because it once again shows that there is an ROI. on the investments and they can even make a positive margin amid this kind of crazy, never before seen growth. My perspective on this is though that their margins will remain very, very volatile, um at least in 2027, twenty twenty seven because they still have this these short term compute demands. That's why they're signing the contracts with um SpaceX. And you could also argue that here from from the kind of midterm elections perspective, with the Democrats likely to slow this build out down, that this is not bullish for Anthropic because compute only got more scarce as the slow as the build build-out out is slowing down and thus the the short short-term term contracts will probably get more expensive for Anthropic
15:41over the next couple of quarters. But the important message from this chart is they have a positive operating margin. And my my strong belief has always been that they would not go public without debt. So they waited until they had a positive operating margin and then they decided to go public. This is, I think, also key to understand. because 79% of their revenue is consumption based. And that is obviously so what you need to understand is they're in a world where they're making long term compute contracts on the procurement side, where they're locked in for many, many years. And then on the other hand, they have a consistent consumption style of business, right? Where they're basically paid per token. Yes, they have a subscription-based part of the company as well, but that's much, much lower versus the um consumption-based part of the business. And if consumption is obviously slowing Going down, then their
16:46revenue would decrease while they're still locked in in the long-term. um compute agreements that they have with with with all the neo clouds and hyperscalers, etc. Right. And that's kind of the the the core risk in their bot business model. Now, as long as they grow as they did, as as I basically showed on the where the second chart, I think, in this deck stack was with their revenue growth. As long as they're they keep growing like that, it's not an issue. But I think if you also listen to Darrow Amoda, Dario Amodi, the CEO of Anthropy, he he he also outlined that as the number one risk that it's super tough for them to decide on how much compute they want to lock in. Um because they need to do that in years ahead because it takes time to build data centers, right? And the business is consumption based. So that's a core risk to understand for also from a margin marginal perspective. right? This I think is also very interesting because what it is showing you is that they are spending about,
17:46I don't know, let's say, I mean, I have the number on the share. So 26% on the the compute they have available is spent on inference, and 47% is used on RD. R D. And what RD means is basically we're using compute to build the models of the future. And the core to understand is that eight more than 80% actually of Q3 revenue came from models that were released in the prior six months. And what and and how you could read that is model development and is is is is core to defend. basically the current sales. And there are two core implications from that. One, there's probably s no slowdown in in in compute demand because they will always need to train their new model. That's that what f that is what frontier AI is, right? You always need the best model, otherwise you will fall behind. And the numbers are proving it, right? Again, 80% of Q Q3 free revenue. came from models that were released in the prior six months. So customers want your Frontier eye, your newest models, at least up until now. And so they
19:01can only monetize, that's what inference means, a certain amount of compute. Now that may increase to a certain extent in in in the future, um, as as as as the quality of the models all um reach certain level and get even higher. But I expect them to to to always keep spending on compute, and that's why they can all only reach a certain margin in their business. I think the the the CEO in their IPO road show uh or the CFO actually told us that they're targeting a thirty 30 five thirty-five 35% percent operating margin, right? And at the moment they're having a 15% margin. Speaking of compute, how are they locking in their contracts? What type of chips are they using? And what are they actually doing to decrease the costs of compute? Because there are certain angles and certain levers that they can pull, pool and they are they they they are focused on that as well.
20:01First, anthropic. in comparison to for instance a SpaceX which relies 100% on NVIDIA or a meta which also relies I think 100% on NVIDIA but I'm on the cheap use but I'm not not fully sure on that or an OpenAI which also has a more than I think 50 or 60% in in OpenAI is that they have a very diversified chip uh supplier base. So they have about 50% from AWS, the training-2 chips. They have Nvidia GPUs, they have the Google TPUs, they have AMD GPUs. And this diversification of the chip supplier base is bringing in many different advantages from my perspective, but the two most important. ones are obviously they can negotiate better terms for future expansion. And here again, the whole China open source and also neo cloud discussion of the Chansen fallout is key because
21:07Nvidia is likely not able to sell to China anymore. And Colette Crest, Cress, the interim CEO of NVIDIA. is outlining that she's not, she was not really a fan of this whole circular financing story, for instance, for the neo clouds. And so that could mean that if she's pulling back on that, and we will hear uh about that tonight, there is the earnings call of NVIDIA, then Nvidia may maybe be losing some pricing power. And this could obviously speak to the advantage of um. a tropic and the frontier AI um uh in general and also SpaceX.
21:43Unidentified speaker (UNKNOWN): Um
21:44Unidentified speaker (speaker_0): so that's the one benefit. The second benefit this and this was uh I think something that I haven't seen much on X but if you read through the IPO statement or the the S1 you in you find in a foot mode basically the they they're basically saying so this is the summary of the footnote, that they were able to get Claude running on AMD chips in just three days with basically no human intervention. Versus in the past, that took engineers, so that took a many, many engineers, and B, it took between six to twelve months to get Claude an entropic antropic model running on new chips, right? And so if they were able to build the models in a way that they can run on very different chips, which they obviously have achieved because they have such a diversified chip supplier base, then this gives them again leverage on price, deliver delivery, and and and and financing. And Obviously, the most expensive exp
22:47part of the entire build-out build out is still buying. the chips, right? Just one number on that for for context. Where a Ruben per gigawatt is costing forty billion. That's so so for you that's forty billion of revenue for NVIDIA. And if they are able now, entropic in that case to diversify between the chip suppliers, supplies. that's a massive leverage they they they get in negotiations with NVIDIA, right? It's
23:14Unidentified speaker (speaker_2): nine o'clock. the kids are finally asleep, the dishes are done, it's time to chill, or maybe start another project, or think about buying a house, or listen to an awesome podcast. Either way, you've got a few minutes to yourself, and researching something like life insurance is probably the last thing you ever want to spend those minutes on. So it ends up on that list for another few months and you put it off again. Soriance, a partner of today's show, is actually built to fix that. At milkrow.soriance.com, milkroad.sariance.com, you can request a free. no obligation quote for term life insurance, and the whole process is online. You can compare options from established carriers and talk to a licensed advocate if you'd like a person to walk you through it. And some policies don't even need a medical exam or blood test depending on eligibility. You've already done the hard part, which is building a life worth protecting. This is the step that makes Make sure it's covered. Get a free no obligation quote at milkroad.sariance.com. milkroad.soriance dot com.
24:01Unidentified speaker (speaker_0): One of the most important charges. charts for me at least in this deck is the CFO of Anthropic did uh a crazy good job. Because if you look at the current balance sheet plus the around about eighty billion that they get from the IPO, depending on where they they at which price they go public now, they have between two 200 hundred to two 215 hundred fifteen uh billion. US dollars on their balance sheets. One of the most cash-rich companies in the entire world at the moment. Um I I looked at the net interesting net interest income in Q3 from their S1, it was $1.2 billion. So that's pretty ridiculous. But the more important part is if you take that the cash that they have against their compute commitments, you quickly come to the realization that at least for the commitments until 2030, they have about 50% already covered. And that is not including any operating cash that they're generating from their business until 2030. So
25:07about half of their compute commitments until 2030, which are I think for yeah, I have it in entitled 451.6 billion.
25:19Unidentified speaker (UNKNOWN): Um
25:20Unidentified speaker (speaker_0): half of that they have already secured, right? That which is pretty crazy. And now it's all about the growth of the company again and how much revenue they can make to make up for the rest, right? But the crazy thing is if you look beyond 2030, right, they have another 450 billion in compute commitment. commitments lined up. That's between 2031 and 2045. So you see that compute is the number one business driver, it's the number one risk. And if you look at that, not from an anthropic-specific lens, but a broader AI investor lens, then at least the demand side um in this trade is covered in this market, is covered. and yes, supply side there there's still many many questions, but I think this was a a major read for for DS1. But I I think you can read this as a as a bullish in a bullish way if you think about investing into a tropic, because yeah, compute is the number one driver and they have a lot of cash to buy
26:22Unidentified speaker (UNKNOWN): to
26:23Unidentified speaker (speaker_0): buy compute. One way for anthropic. to reduce the costs is to think about basically having your own data centers. And that's what they're doing with uh something called Anthropic Direct. At the moment actually all of their compute is rented from NeoClouds, from hyperscalers, from from SpaceX. And with Anthropic Direct, what they're trying to do is they try to sign long-term leases of the shells, basically, of empty shells, so empty data centers, while at the same time, Anthropic goes out to its diversified chip supplier base, buys the chips, supplies the companies that are operating the data center shells Else that Anthropic has leased, least, and and then run it. And what this does is it gives them flexibility in terms of which chips they need to secure. Um it and in let's say 2028 and beyond, and maybe also a bit earlier, it it takes away their need to rely on short-term capacity, right? Depending on how much Anthropic direct, how much how many gigawatts of Anthropic Direct
27:40will
27:41Unidentified speaker (UNKNOWN): um
27:41Unidentified speaker (speaker_0): eventually cover, right? And so and and because what you need to understand is the expensive part in the entire data center part is buying the chips, but the scarce part at the moment is getting the powered shells, the powered data center buildings, the power, the empty. data center buildings that have power available. And so they rent them earlier while they can later decide on which chips they buy and then get those data centers operational. And I think that's a super smart kind of procurement approach to get less dependent on those short short-term term agreements that they have with suppliers like SpaceX. to reduce compute costs eventually and then to also increase their operating margin, which is key, of course. What are we doing with Entropic? Are we buying the IPO or not? Let's look at the number one question that I think you, as an investor, need to answer before buying this stock. You need
28:48to believe in the long-term growth of this. company. And this is actually becoming the 600 to 700 billion ARR company that Goldman Sachs, for instance, is forecasting by 2030. Because obviously, growth is covering everything here and is the entire investment story. How can they grow? I I I believe in three core levers for from a short more short-term perspective. I think it's them covering the kind of daily recurring work in operations, in finance, in security. What that means is that more and more AI will be adopted in more and more company processes and more and more um processes at work will be automated with AI. And this is kind of the short. short-term lever that they have. Midterm, it's more from an agentic perspective that they're earning with the adoption of other products. So think about today you have, for instance, Microsoft Copilot, which where you can select, you use the OpenAI models as a user, as a copilot user, you can choose which model you use, right? In the future, companies like Salesforce, right? They're already doing that, will offer their agentic solution that they build on Claude.
30:07And obviously, the more those agentic use cases of the partners like Salesforce, maybe also Copilot, will be used in the companies, the more revenue can uh Entropic make. But the biggest long-term lever, in my opinion, is I call it here Claude being part of the of of future RD budgets, right? Claude in science, Claude as industry tools, where where in the end the AI tools of Anthropic are providing finished work, finished deliverables. And maybe they even get priced Price per output, not per token anymore. But this is where I see the biggest upside in the future. This is where, of course, companies can replace full um FTs, so employees, maybe also entire teams, because AI can just deliver the final work in a reliable way. And so those are the three different growth drivers that I see. Now, if you compare Anthropic from a valuation perspective on price to sales, it's actually
31:15not that expensive if you compare it to an OpenAI at the latest private round or a SpaceX at the current market value. I think that's of today or yesterday 1.87 trillion. Um at the $74 midpoint, um, so that's IPO price for anthropic, you're paying about 8.2 price to sales for. their 2027 estimate sales, which is fair, I guess, if you look at their growth. Now, if you look at their multiples from a free cash flow perspective, from an earnings perspective, obviously this the discussion is a is a completely different one. In the end, everybody for themselves needs to make the the judgment call, the investor call, on whether you think this is expensive. fairly priced, or even cheap. And to me, it comes down to a revenue growth and b revenue per dollar of compute that they're spending, and that they can defend that and and and eventually it needs to grow.
32:21Support this in kind of stalled, right? From two 2.0 point something to one 1.72. point seven two, uh, but that needs to grow. Me personally, and that's the last slide I have with me. I'm not buying DO IPO.
32:33Unidentified speaker (UNKNOWN): Um,
32:34Unidentified speaker (speaker_0): to me, I think it's a pretty aggressive valuation, it's a pretty high valuation. Um, and there are four core reasons for that. Number one, and this is where Dario Amoda, Ahmudai, I think, agrees with me: the compute commitments are the biggest risk. This could question in or answering the question of how much risk are we taking in allocating and securing compute today, versus how much demand do we have in the future, and if we allocate too much today, are we running into bankruptcy or not? Or at least running into very, very harsh margin headwinds. And obviously, if you're a public company, then your stock goes down a lot. And if I just look into 2027 with the midterm selections, with the slowdown of the AI data center billot, we remain in a very compute-constrained environment. Yes, they have entropic direct, but this is only scaling and as has is basically providing a very small share of compute that they're using today.
33:41Then I just think a company like SpaceX is in a better negotiation position with another pic versus the other way around. And that's why I think that's number one risk over the short term and why I stay out for now and at these levels. Number two, sovereign AI is something Palantir is talking about a lot, something Palantir. Eli Lilly, for instance, is also doing. Basically means that
34:07Unidentified speaker (UNKNOWN): it
34:07Unidentified speaker (speaker_0): means two things. On the one hand, companies are building their own models, um, like Eli Lilly. They have the tune lab, right? And why they're doing that, they're doing that because they do not want to give away their core company um IP, their the core secrets that that that make up the entire value. of the company right and I still think this is a major headwind me being in a European company I s I I see that in the board discussions as well that this is this is a headwind that that people are not trusting the Frontier eye with with putting it into the core um kind of kind of environments. advice. Now I think this is only true to a certain extent because the big consultancies, the big banks, um, also Eli Lilly for the kind of routine work has adopted um Claude, but it is a headwind in when it comes to to AI adoption, of course. And then number three and four, and I think they're they're they're correlated
35:07this whole commodization arguments, as there there's are so many peers and suppliers of AI tools out there right think about the frontier think about frontier AI right the there's not only entropic but you have OpenAI then you have all the open source um models that are out there not only from China but also from the US and um
35:27Unidentified speaker (UNKNOWN): and
35:27Unidentified speaker (speaker_0): there's the one from Palantir and Nvidia I think nemotron is the name then you have all the agentic tools like Muse like Rockbot like Dots and so this will remain a very, very competitive market with pricing towards zero. And yes, I think you have a shot at monetizing frontier AI, but for that, you need to remain at the frontier, you need to remain um uh heavy on the gas when it comes to compute spending. And I just see a risk that the market will price that down. eventually. And yeah, and and and taking those risks together, I'm not buying the IPO. I understand everyone that is buying into the IPO because A, it is b much below the two trillion dollar valuation, which I think is super smart because it enables entropy to ru to to to raise more cash. And I mean the growth numbers of the past they speak for themselves, right? Um And it and if they were if they're able to keep growing like that, then the valuation is probably fair and there is upside in the stock. Um but I just want to remain I I usually don't buy into IPOs. I did not buy into the
36:36SpaceX IPO directly. I just want for the hype to kind of go away for for X to not talk at least as much about entropic Enthropic and into their IPO as as it currently does, and then for the stock price to come to come down a bit. I could easily envision this stock price to rise a lot on IPO day on Thursday. And that's what I'm waiting for. And then we'll see whether I buy into it or not. And uh yeah, with that, I wish you all the best. Let me know. I'm really curious if you guys are buying the the the anthropic IPO or not. Let me know in the comments and uh yeah have a great evening.
37:23Unidentified speaker (speaker_2): Thanks for listening to Milk Road. If you enjoyed the show make sure you like and subscribe. And if you're struggling to find winners in the market, that's exactly what MilkRoad Pro is built for. Our analysts have called some of the biggest winners early and Pro lets you see what they're buying next, every trade they make and the research behind everything every position. Check out Milk Road Pro at the link below. Everything you hear on Milk Road is for informational purposes only. These are our personal opinions, not financial advice. And we may own some of the investments we talk about. Always do your own research and make the decisions that are right for you. See you next time.