Anthropic's Historic IPO Has Arrived | S1 Breakdown

0:00It is here, folks. This is the Anthropic IPO S1 breakdown. We've been waiting for this to drop for, I don't know, feels like forever. This is my Super Bowl. I'm going to take you through the fastest growing company with the most mind blowing efficiency metrics I've ever seen. The IPO ever. Get ready for the Anthropic S1 breakdown. Is this thing on? Yesterday's price is not today's price.

0:35The Anthropic S1 dropped yesterday morning and it was such a momentous occasion that my wife made me pancakes. So we've been together for 13 years and the number of times she's made me pancakes I can count on two hands. So it was a glorious and delicious human moment as I set off to work to cover a human-like technology that will either zap us from humanity or cure cancer and help us all live until 200. For the record, we won't be touching the third rail in this piece as to whether our existence is threatened. I'm bringing this syrup today. Pearl Mill and Company, formerly Aunt Jemima. Anthropic filed to list in the NASDAQ under the ticker ANT.

1:13The price range is still blank, but bankers have told investors a deal could raise$100 billion or more at a valuation near$2 trillion. And that's according to the New York Times. So the last private round was$65.2 billion raised during their Series H that closed in June at a reported$965 billion valuation. So a little bit more than twice as much as that last valuation in the private market. So they say, for the first time in history, intelligence is something humans can build. That is a scary and motivating sentence. Anthropic did$19.7 billion of revenue over the trailing 12 months, ended June 30, 2026.

1:51The second quarter alone was$11.6 billion, which was up, wait for this, 1 ,368 % year over year. The three biggest questions from my PPOV, that is Pancake point of view, is number one, can you be the most impressive technology to ever exist and also a commodity where your valuation gets driven down? Two, will the use cases keep growing as fast as the model costs are dropping? And three, is there such a thing as having too many partners at the party? Key stats. Revenue,$16.3 billion for the first six months of 2026, up 1 ,300 % year on year. That's up from$1.2 billion in the first half of 2025.

2:31Q2 2026 was$11.6 billion, up 144 % from Q1. That revenue is blank in the filing, even though September has come and gone. They will no doubt have that in before they IPO. Revenue mix, 80 % coming from consumption, 20 % from subscription. So consumption, which they call pay for intelligence, is metered in tokens and price by model. It was$13.1 billion of the first half,$16 and change billion in revenue. Subscriptions were$3.2 billion, up from 17 % of revenue in 2025. This is the classic 80-20 folks. Customers paying$100 ,000 or more. They have$11 ,900. Wow. That's up from more than$950 a year earlier.

3:13and more than 1 ,900 pay 1 million bucks or more and about 250 pay 10 million or more. Another 40 million customers each pay less than 100 ,000. Okay, here's a spicy one. Net revenue retention, 900%. So this is measured as this quarter's revenue from customers who spent at least a dollar in the same quarter a year ago and pro and max subscribers are excluded. So your boy wasn't in that count, but I like to use it, you know? Customer concentration. no customer over 10 % of revenue for the nine months ended September 30th. About 79 % of Q3 revenue came from outside the top five customers, which was up from 64 % in Q1.

3:53Revenue from each of the top five still hit a record in Q3 despite that mix shift. But 48 % of first half revenue came through the cloud marketplaces that was up from 11 % in 2023, representing 58 % of receivables. shout out to all my folks and accounts receivables holding down the fort operating income 974 million in q2 and eight percent margin well that's a positive number what are we seeing here the only quarter in the filing with an operating profit good quarter to go public right the first half overall was still a 693 million dollar operating loss adjusted operating income 2.2 billion in q2 a 19 % margin,$1.6 billion for the first half, a 10 % margin.

4:38This adds back stock-based comp and non-cash charitable donations, plus the$1.5 billion from that BART settlement for copyright shenanigans in 2025. Can't do that. Net loss is$80.7 billion for the first half. $72 billion of that is from marking up convertible notes as the valuation went up, and$8.8 billion is mostly a fair value charge, a note purchase agreement. So almost none of that is non-cash, right? The valuation is just going up and it hits the P &L. Adjusted net income was a positive$2.7 billion. Adjusted. Very adjusted. Compute and infrastructure. Who's footing the bill for this? My goodness.

5:16$8.9 billion for the first half. 55 % of revenue. That was down from 201 % of revenue a year earlier. Q2 alone was 48%. Q2 revenue grew 1 ,368 % year on year, while compute grew 300%. Gross margin, about 85 % in the first half. If you back into it, back up the bus. They switched the income statement to expense types in 2025. I caught you. But the footnotes still break out. Cost of revenue,$2.4 billion. Again,$16.3 billion of revenue. So up from 40 % in 2025 and 95 % in 2024. Free tier compute and channel partner fees sit in sales and marketing. It's like their cost of going to market and acquiring customers, even though it's typically in COGS.

5:58So the fully loaded number is lower. More on this later because it ain't 85%, even though you can kind of see it might be 85%. It's not 85%. Stock comp, 2.2 billion in the first half. Must be good to be buying a house in San Francisco right now. Another 2.9 billion gets recognized at the IPO, then 23.7 billion. My goodness. Over about 2.1 years plus 3.1 billion on other awards. I'm saying, oh my goodness, as if I didn't write this before. I don't know. Every time, you know, sometimes I surprise myself in my own writing. This will continue to throw off the casual investor when they have to see such a large loss that is due to their success in totally non-cash.

6:37Interest income. This is just funny to me. $1.1 billion in interest income in the first half because they still have so much cash. $762 million of that was in Q2. So this is larger than most public companies' full-year revenue that I cover week to week in my newsletter. Okay. So in six months, they made$1.1 billion in interest revenue. Just hilarious. Operating cash flow,$3.9 billion in the first half compared to$1.5 billion a year earlier. CapEx was only$482 million because Compute has mostly been financed through agreements with the cloud providers. Is that what you call payment in kind? Not getting into that, accounting treatment.

7:11Cash,$131.7 billion at June 30th. $112 billion of cash plus$19 billion of short-term investments. They've raised$141 billion to date. Compute commitments. They kind of left this blank in here. Aggregate committed spend of more than blank running through 2045. How are you going to leave us hanging like that? Employees, 5 ,300, up from 940 at the end of 2024. Imagine being in that Slack channel like who are all these people? TTM revenue per employee is over 6.6 million, up from 1.4 million a year earlier. That is crazy. The offering, the price range and share count are blank. Goldman, Morgan Stanley are joint representatives, then J.B.

7:53Morgan, Citi Barclays, and Wells Fargo, and another 22 banks. It's a controlled company with five classes of common stock, which we will go over. And they did it 10 for one split on September 17th. I'm going to level with you. This was the best shareholder letter I've ever read. And I have to admit, even if it was written by Claude, but I don't really know, this thing slaps. It goes from, we will try to cure cancer, to here's why you might not want to buy this, to an explanation of what exponential really looks like and why they're going to make some investments that look, quote, irresponsibly large on paper, but are justified in this new world.

8:28This letter really had it all. Quote, we believe that AI is one of the most consequential technologies in human history. If managed well, it could rapidly accelerate scientific innovation, greatly increase economic growth rates, bring cures for major diseases within reach, and solve problems that have resisted generations of human effort. They lead with the scientific angle, cure diseases and extend life. And I think that's very thoughtful positioning versus potentially killing us all. And they name drop a bunch of diseases that they hope to cure, which is dope. That's awesome. Then they hit you over the head with the four things they want any investor to understand.

9:04The potential of AI, the risks, how anthropic is built differently, and the ways in which our behavior may differ from that of a purely profit maximizing company. Quote, we believe that the impact of an exponential and machine cognition will be vastly greater even than that of the exponential in computer chips. Where it ends up is what they call a country of geniuses in a data center. Millions of virtual agents with the ability to revolutionize every area of human endeavor. And boy, oh boy, is that a weird ass statement to put in there. They hit us with this anti-cell. Before starting the IPO process, we spoke to many business leaders who had gone through a similar process in the past.

9:42One piece of advice, which we heard in slightly different forms from multiple people stuck with us. Never let your company be worth a dollar more than it is actually worth. But like, also value us at Qtroy. What they're saying here is the same thing that they do in recruiting, where candidates get like this anti-sell pitch of like, it really sucks to work here. You do not want to work at Anthropic. Then they warn you of surprises coming down the pike. So if it first would be spending, business bets that derive from AI's wild exponential trajectory but seem strange or unorthodox in isolation. And on the public good side, decisions that depart from profit maximization in order to appropriately balance the public good.

10:22Quote, Anthropic will often make forward investments, especially in compute, that may seem irresponsibly large at first glance. air cover, they compare it to pharma where you get these huge upfront R &D investments and they only make sense later in light of the returns from successful drugs. Except for Anthropic, it's a similar process that's occurring much faster and on a much larger scale. Their stated approach is to, quote, be bold but sane. They've also committed to keeping Claude free of advertising. That's a weird choice to say we're not going to do this one major business model that has funded a ton of the innovation in the economy.

11:00And they say there may be time that they choose to advance more slowly than they could, even if that puts us at a competitive disadvantage. They sign off asking you as a prospective owner to quote, not to think of your purchase in purely transactional terms. I left thinking I would buy a used Chevy Avalanche from them. A refresher where Anthropics sits in the AI stack. So if you read our N-scale S1 breakdown, you've seen the five layers of AI. So first you got power. So that's the electricity itself from the grid or generated on site, or maybe nuclear. Then you got land and data centers, the buildings, the cooling, the grid connections that turn power into usable space for servers.

11:38Then you get the chips, the GPUs, and other accelerators, plus the networking gear that ties them all together, all those bungee cords. Then you have cloud software, the layer that schedules jobs and keeps the hardware running. And five, the models and apps. So you get Anthropic, OpenAI, and everyone building on top of them. So Anthropic lives at layer five. It rents most layers one through four from the Amazons, the Googles and the Microsofts of the world. And it trains in service cloud across four chip families. You got Amazons, Googles, NVIDIAs and AMDs. But it's starting to reach down the stack.

12:10And what I mean is they're going to vertically integrate here. They call it Anthropic Direct, leasing chips directly and hosting them in dedicated facilities operated on their behalf, which now puts equipment leases and committed power costs on their own P &L. The filing also mentions guarantees to developers to get data capacity built. Quote, as we continue to build out our own computing infrastructure and further vertically integrate across the compute supply chain. Remember when pricing was simple? One product, one subscription, one invoice every month. Those days are over. The AI economy is changing how everything gets bought and sold.

12:46We're seeing usage-based pricing, hybrid contracts, and something called a credit. The finance systems built for vanilla subscriptions weren't designed for this. This isn't just a RevRec problem anymore, because outdated revenue architecture can kill a great pricing idea before it ever reaches the market. RightRev lets you recognize revenue in whatever shape it comes in. Whether you're launching consumption credits or testing entirely new commercial models, RightRev gives finance the flexibility to support whatever pricing model comes next. So product teams can keep on innovating, and finance teams can keep pace.

13:19If you're architecting revenue for the AI economy, learn how RightRev can help. Visit rightrev.com slash CJ. That's me. That's rightrev.com slash CJ. A wise man once told me that every startup is really just a loosely functioning disaster. Ain't that the truth? If your startup needs a little extra support, and let's be honest, who doesn't need a little help now and then? You're in for a pleasant surprise with Fidelity. Fidelity Private Shares helps early and growth stage companies stay investor ready with cap table, data room, and scenario modeling all in one place. A messy or missing cap table might not just slow you down.

13:56It could cost you your next fundraising round. VCs are flooded with pitches and if your equity is confusing or missing, they'll move on fast. Fidelity Private Shares gives founders the structure and simplicity to focus on what actually matters, building your company. If you stay investor ready, you don't have to get investor ready. Check out fidelity FidelityPrivateShares.com to learn more. That's FidelityPrivateShares.com. I got news for you. The ERP category is finally getting disrupted. And if you haven't heard of it yet, please pay attention. It's the AI native ERP built specifically to replace NetSuite.

14:31And it's already won over hundreds of finance teams. Their mission is to make the zero-day close a reality. And they're actually doing it. We're talking teams closing the books at 1.35 PM on the first day of the month. Companies like Windsurf, Mercore, and hundreds of others run on their entire finance stack on Rillit. Revenue recognition, close management, multi-entity, native Stripe and Salesforce integrations. Woo! Everything a scaling company needs. They've got 5.0 stars. 5.0, wow. 5.0 stars on G2. They're backed by A16Z and Sequoia. Heard of them. And CPA-led implementations that get you live in 45 days.

15:06That is simply unheard of in the ERP space. If your books aren't running as fast as your business, check out Rillit. Book a demo at Rillet.com slash CJ. That is R-I-L-L-E-T dot com slash CJ. That's me. How are they making money? They're making a lot of this money. There are four model classes. They got the Mythos, Mythos, Mythos, including Fable, Opus, Sonnet, and Haiku. Each one sits at a different point on what they call, I love this, the Capability Cost Speed Curve. Customers use the smart, expensive ones for coding and agents and route the simple shit. to the cheap fast ones. I forgot my password.

15:44I forgot my login. They charge two ways. The first is pay for intelligence. Customers pay for usage metered in tokens and price by model. This was 80 % of revenue in the first half of 2026. And subscriptions, clods apps on a monthly or annual plan for enterprise and prosumers like myself with terms of 12 months or less. This was the other 20%. The two models feed each other because when a subscriber runs out of the usage in their tier, they can move up a tier or buy more intelligence directly, which turns that subscription revenue into usage revenue. And that's why the usage revenue is far outpacing the subscription revenue.

16:17A growing number of enterprises also sign non-cancelable annual or multi-year contracts with minimum spend commitments in exchange for a committed price list. So like to lock in with certainty what you're going to pay. These mostly show up as usage revenue too. So they sell it through what they call surfaces. Cloud Platform, the building blocks of developers to put Cloud inside their own products. Cloud Code, the coding agent. Cloud Cowork, the knowledge work agent, which they started merging with Cloud.ai in September. Let's talk about distribution, which they called diffusion. Much of the diffusion of Cloud has been organic because prosumers use it in their own personal life and then eventually they bring it to their company life.

16:56And by the time the sales team from Cloud is there, they're already using it in some way, shape or form. They're looking to get better pricing. The biggest benefit to the go-to market is that it's available through all three major cloud providers. So a customer can buy it inside the AWS, Google Cloud, or Azure relationship they already have. And Anthropic keeps the contract with the customer and pays each cloud a fee, a vig, a rake. And then there's the old-fashioned channel model, the consultants, the Accentures, the Deloitte's, the KPMG's and PWC's of the world. They all have these cloud practices.

17:29They will make a ton of money in this boom, just implementing it. And then Salesforce and Snowflake embedded into their platforms as well. So a lot of ways to buy. They also see the private equity angle. PE's eating everything, baby. The problem is how you get it into midsize companies who aren't that tech forward. So they launched Ode with Anthropic, that's O-D-E, an AI services company built with PE firms to deploy cloud across their portfolio. So sales, marketing, and partnerships came to 19 % of that massive revenue in the first half. And most of that was fees to channel partners. They don't call it ARR.

18:03So a few weeks ago, I wrote, there are no new crimes, just new ways to calculate revenue. Ain't that true? Well, after Higgs Field's CEO was on 28 Minute VC, and he said that they calculated annualized revenue, quote, the same way OpenAI and Anthropic do, by taking the last 28 days of revenue and multiplying by 13. Anthropic doesn't call it ARR. They call it run rate revenue. Quote, the sum of one, our consumption-based fees during the trailing 28 period ended on the measurement date multiplied by 13. So I think that's like 364. They give you one day back. How generous. And two, our aggregate monthly subscription fees as of the measurement date multiplied by 12.

18:41Like I said, last 28 days times 13 plus monthly subscription times 12. But this is the kicker. They don't give you a number. They just define it. That is a first for me to have this definition and then not give us a number. However, they've been telling everybody and their mother for months. So Dario said on stage in April that they'd hit a 30 billion run rate. Then the series each announcement in May said 47 billion. and Bloomberg reported$65 billion at the end of July, up from about$9 billion at the end of 2025. When the$30 billion number came out, OpenAI argued, hey, hey, hey, not the same, overstated by about$8 billion.

19:14And the filing shows why they might say that and be right. Because Anthropic books cloud marketplace sales at the gross larger amount before the cloud providers take their cut. As for year end, the Financial Times reported investors expect a run rate to land somewhere between$100 billion and$120 billion at the end of the year. So keep that in mind for the valuation portion. If I'm being honest, I like the old school version of run rate revenue better, where you take your last quarter and multiply it by four because you have three months in there to smooth any sort of seasonality out. So if you do that with Q2, you get around$46 billion, which is right in line with that $47 billion they announced in May, and the roughly$16.7 billion in Q3.

19:53I can back into that. I'll show you in a minute how I do the wizardry, do the math in here. And if you multiply that out, you get the$67 billion or close to the$65 billion that Bloomberg reported in July. Net-net, I think gap revenue is a much more honest view of what they can actually do than this run rate kabuki. Usage revenue moves around a lot. For example, during the two weeks ended January 2nd, they said total customer spend dropped about 20 % compared to the two weeks before because developer commits slowdown during holidays and all the retail companies do code freezes. Quote, run rate revenue is not a substitute for revenue prepared in accordance with GAAP or a forecast of revenue.

20:32I feel like they listened to me. Did I influence them? They also say recent run rate growth was driven in part by rapid initial customer adoption and expansion, often characteristic of an early adoption period, and that they expect the growth rate to decline possibly significantly and more rapidly than we anticipate. I still bet dollars to donuts that we start hearing a lot more about 90-day averages when that starts to slow down. You can go back and forth on revenue growth expectations, but you can't argue with the revenue efficiency. These are the highest efficiency numbers I have ever seen for a company.

21:03Net revenue retention is about 900%, meaning customers who are spending a year ago are now spending about nine times as much. Snowflake was at 158 % when it filed, and GitLab was at 152%. People thought that those numbers were crazy. This is multiples more. Some of this is admittedly a feature. It's not a bug because of their pricing model. So AI is pretty much synonymous with usage-based pricing at this point. And 80 % of Anthropics revenue is consumption. So revenue go up as long as customers use more. While I don't think this figure will stay at 900 % forever, even if it is cut to 300%, that would be S tier.

21:38Also, it's crazy how they can both have 900 % NDR and also be a commodity product with little to no switching costs. How can both of those things be true? Oh, and revenue per employee, 6.8 million. I would love to have that in my company, up from$1.4 million a year ago. My rule of thumb for a public company is$450K to$500K is the median. TTM revenue here grew about 1 ,200 % over the past year, while headcount growth grew about 170 % from roughly$940K to the$5 ,300K today. I know I just said that everyone was talking about revenue too much, but you can kind of back into the revenue as of September, which they don't disclose using the revenue per employee in the employee count.

22:20They say TTM revenue per employee was over$6.6 million as of September 30th, which is approximately 5 ,300 people working there, which puts trailing 12-month revenue at roughly$35 billion. So if you subtract the three quarters, we do have Q425 through Q226. About$18.3 billion comes out. And Q326 comes out to roughly$16.7 billion, up from$11.6 billion in Q2. I love math. I want to take you through four drivers because they interestingly have four different time horizons. Okay, so the filing lays out the company's financial model. It's important to outline these four. So one, adoption of Claude.

22:59Each new model version brings in new customers and pushes existing ones into new workflows and revenue growth. Two, compute efficiency. More intelligence and more revenue out of every dollar of compute. Three, operating efficiency. Talent density and efficient go to market in their own teams using Claude, which shows up in operating margin and revenue per employee. Fourth, compute investments. The big checks. So the first three have all been moving up very, very fast. But the fourth one that I mentioned, the compute investment, that lags. So they say compute is procured and committed in advance and the expense is incurred as it is brought online and added to our fungible pool, often in large, discrete steps.

23:37They typically commit to compute 12 to 18-ish months before it's available. And once it comes online, the expense hits the P &L whether or not the revenue it's meant to support has shown up yet. So every time a big tranche lands, margins will take a step down. And they tell us that directly. They say, we expect operating margin, adjusted operating margin may compress in those periods. So when demand outruns their long-term capacity, they also buy short-term compute, which costs more per unit and they expect to lean on more of it in the near term. Here's capital allocation for Compute 101. They say, quote, we manage compute as a single fungible pool for multiple strategic partners and we dynamically direct available compute between workloads based on where the marginal unit is expected to be most productive at any given time.

24:26Net net, it's one pool covering research, model development, serving customers and internal use of Claude. They're using their resources for the best marginal use case. By the way, revenue per dollar of compute, I think it's a really cool stat that companies should try to adopt internally. Their headline version is that Q2 revenue grew 1 ,368 % year over year, while compute and infrastructure expense grew 300%. If you run it quarterly off their P &L, you can kind of see the lag. And that's a 5.2x improvement in compute efficiency over five quarters. Because yeah, why not? Give us a weird ass number of periods, but it looks great.

25:02And to my point above, in Q425, compute expense jumped 44 % from the prior quarter as new capacity came online. Then revenue caught up, and in the next two quarters, it doubled, then tripled. So it's kind of like slow, slow, then everything at once. I know you've been waiting for this question. Are they actually profitable? In my revenue crimes piece I wrote, I asked whether they were still selling a buck for$0.60. Or maybe, hey, maybe it's$1 for$0.75 now. In Q2, they spent$10.6 billion of total operating expenses to bring in$11.6 billion of revenue, which actually works out to about$0.92 per dollar.

25:39$974 million of operating income. That's an 8 % margin. And it was actually the only profitable quarter in the filing. But if you zoom out to the first full half, they are still$693 million in the whole. And every full year, by the way, before that, big losses too. So they got the results from operations in there. It's interesting if you take a look. And the$80.7 billion net loss of the first half is kind of noise from a cash perspective because$72.4 billion of it is marking up convertible notes because the company got more awesome and more valuable. Another$8.8 billion is mostly fair value charges in the note purchase agreements with their partners like the Amazons and Googles of the world.

26:18Their stake is worth more too. So adjusted net income was a positive$2.7 billion. if you adjust, adjust, adjust. It's the same thing I pointed out when the 2025 numbers leaked when they had this$34 billion of the$42 billion of loss that was non-cash revaluation of financing. If there is a gross margin, I will find it. So they pulled gross margin off the income statement last year when they switched to grouping expenses by type. They got it now bucketed by compute, personnel, sales, and marketing. But the footnotes still break out cost of revenue the old way. So in the first half, it was$2.4 billion.

26:52Again,$16.3 billion of revenue, which is about 85 % gross margin, up from only 40 % in 2025 and negative 95 % in 2024. Two major caveats. The free tier compute and the cloud partners cut both sit in sales and marketing. So it's got to be lower than 85 % because channel partner fees went up about$2.7 billion year on year in the first half. And the compute behind free usage is somewhere inside the remaining sales and marketing line. Okay. That's how they acquire their customers by giving away their product for free, which caps it at roughly 600 million. So if you load those back in, it's more like the high sixties, maybe low sixties.

27:36And since they book cloud marketplace sales gross, the top line includes the cut that goes to the cloud providers. So the first half is closer to 13.6 billion than 16.3 billion net of it. So either way, if you book it as an expense in COGS or you just reduce the revenue, it's the same circle to square there. You are what your gross margin says you are. Sorry, my kids are tired of hearing me yell that at the kitchen table. Dad, I'm two. I don't know what cloud costs are. Well, I do. And cloud infrastructure is the second largest line on most tech P &Ls. And yet, most finance leaders couldn't tell you how it's priced.

28:11And as a CFO, I understand there's a non-linear impact to your company's valuation when you improve your gross margin. and the path of least resistance is through your AWS or GCP bill. Because like anything you buy, cloud gets cheaper when you buy it in bulk. In fact, you can get the rate down by 20%, 30%, even 60 % in some cases. The flip side is getting saddled with a bunch of capacity you never touch. Happened to me once, I think I still got some megawatts in the kitchen cupboard somewhere. That's exactly what cloud capital solves. Cloud capital underwrites the risk so you can realize the same savings of a deep, long-term commitment sands the downside.

28:49Oh, and it takes no engineering work. Bingo! I've known Edward since back when he was just starting the biz, and it's incredible how much they've saved customers on their cloud bills. That could be you too. So head over to cloudcapital.co, that's cloudcapital.co, and find out how much you could save. Every decision a company makes is inherently a financial decision. Folks, I've been trying to tell you that for 300 episodes. So it's wild that most finance teams have to spend a week on what already happened. While my favorite movie is Back to the Future, we'd be living in the past. Speaking of movies, here's a terrible plot.

29:25Tired accountant chases down a receipt from March and then polices a$20 Anthropics subscription, starring the accountant Ben Affleck. So stop living it. You don't have to play defense on what already happened if you get your system sorted. And Brex's intelligent finance platform clears the backward-looking work. policy enforced at the point of sale receipts matched instantly close short into days not weeks what's left is the forward-looking part where to deploy capital how fast you can move what next quarter actually needs because the defense is handled by brex it's time for finance to go play offense throw the ball deep thousands of companies including anthropic coinbase and doordash already run on Brax.

30:08Head to brax.com slash metrics to join them. That's brax.com slash metrics. If you work in finance, you know what time of year it is. I'll give you a hint. It's our Super Bowl. Budget season is here. Woohoo! So how many versions do you think you'll get up to this year? I once got to V147 in the good old Excel days. Beat that. Actually, please don't. This is when your planning software should earn its keep. You don't have to choose between either Excel or an inflexible ballooning platform that takes a consultant to change. Abacom brings you the dynamic capabilities you love from Excel while preserving the rigor and governance of an enterprise suite.

30:46I met Abacom's founder and CEO, Julio Martinez, more than three years ago. He and his co-founder, Jorge, came out of finance and they built a planning platform that can run a complex business without becoming another system that requires a babysitter. Abacom puts your actuals, forecasts, models, and business context in one place. So you can update the plan, run scenarios, and answer why did this change. In fact, you can even ask it in plain English. You might not even have to spell it right thanks to AI. That's why I'm so excited about Abacum. Go to abacum.ai slash mostlymetrics. That's A-B-A-C-U-M dot A-I slash mostlymetrics.

31:26Or just tell Julio that CJ sent you. That'll probably work too. Cash is getting better. They generated$3.9 billion of operating cash flow in the first half compared to burning$1.5 billion a year earlier. And CapEx was under half a billion. So call it about$3.4 billion of free cash flow. Some of that is work in capital timing since they collected cash up front from customers and took longer to pay vendors, partly offset by a$5 billion jump in receivables. Wouldn't that be nice? And keep in mind here, the stock comp hasn't hit yet. If you read our Aura ring breakdown, Now, and they didn't end up IPO-ing, but I still did it.

31:59This is the same setup. So most of their RSUs need two triggers to vest, which are time served and a liquidity event. And until the IPO is effective, they can't book the expense because the IPO isn't considered probable. Quote, if this offering had been completed on June 30th, 2026, we would have recorded 2.9 billion of cumulative stock-based compensation expense related to the RSUs on that date in an additional$23.7 billion of unrecognized stock-based compensation expense related to the RSUs, which would be recognized over a weighted average period of approximately 2.1 years. Oh, and there's another$3.1 billion of other awards on top of that, and they're using the accelerated method, which front loads it.

32:40Part of the IPO proceeds is also going to cover that tax bill when those RSUs settle. So TLDR, I know I gave you a lot there. Are they profitable? tell him what he's won johnny for one quarter yes with 974 million of operating incoming q2 cash looks better at about 3.4 billion of free cash flow in the first half but they're telling you margins will dip every time a big chunk of new compute comes online and the ipo kicks off 23.7 billion of stock comp over about two years which will hit the pnl that's roughly 11 billion a year against an operating profit running around 4 billion makes money on cash today but gap probably goes back into the red until revenue catches up.

33:18Will the use cases keep on growing as fast as the costs are dropping? Let's check out the flywheel they got going here because this part is rather counterintuitive. Claude Opus 5.5 cut customer cost per task by about 40 % compared to Opus 5 and their argument is that cheaper tasks make people use more of it not less. Quote lower cost per task drive greater usage which in turn drives broader diffusion of Claude. Our compute efficiency allows us to deliver these reductions while preserving our margins. My number one question is whether the use cases keep growing as fast as the model costs are dropping.

33:55Adoption curves are never an even path and they're buying compute 12 to 18 months ahead of a fixed cost. So if prices fall faster than people find new shit to do with it, the compute build does not shrink with the revenue. Optimization we have found waits for no man. You got to ask yourself if use cases keep growing once we hit a point where the models are just smart enough for everything we do today. I mean, hell, we need to actually start inventing stuff to do with AI. So to their credit, the letter says that, hey, that's the plan and they'll spend less effort protecting what they already sell.

34:28Quote, AI chatbots became viable around 2022 and AI coding agents around 2025. AI agents for general knowledge, work and scientific discovery are gradually becoming viable in 2026. So you will have a lot more science use cases. They launched 250 new model versions, platform features and applications in the first nine months of 2026. And models released in the prior six months made up over 80 % of Q3 revenue. So customers move over to whatever is new almost immediately. They want to drive the Ferrari, right? Nobody wants yesterday's newspaper. The flip side is that the newest stuff is agentic and agents chew through much more compute per interaction than a simple chat.

35:06And the risk factor says agentic work grows. the compute needed to serve customers, quote, could increase disproportionately relative to the associated revenue. So a weird question. Is it actually in their favor to not let the models drop too much at a rate faster than we can diffuse them into our lives? How much wallet share is left of these big customers? So over 50 % of the Fortune 500 already spent$100 ,000 or more a year with them, but that spend is less than, this is crazy, 0.1 % of those companies' total expenses on average. Even the top 10 Fortune 500 companies spend less than 1 % of their total expenses on Anthropic, which is still eight times the average.

35:46That was kind of mind-blowing to me. Is there such a thing as having too many people at the party here? Like Amazon, Google, Broadcom, their investors, their suppliers, their lenders, and the filing spells out how tangled it gets. They say, quote, Amazon, Google, and Microsoft, through their respective affiliates, have made significant equity investments in our company. These CSPs are also our customers, purchasing our intelligence service for internal use for their own products and services. So they're also competitors, by the way, since all three sell their own AI models on the same clouds that resell Claude.

36:19The filing says that being on their platforms gives them visibility into our pricing in commercial terms. As far as counterparties go, there are large notes to Amazon, Broadcom, Google, and Amazon. They hold billions of dollars of convertible notes. I mean, you'd normally want to diversify your counterparty risk, but like when does having too many counterparties hurt you with the others? Or do they just not give a shit because Anthropica is the biggest counterparty? Does it get to the point where if you work with everybody, you can do whatever to anyone? It's both leverage and cost because 48 % of first half revenue ran through these cloud marketplaces, which also makes up 58 % of receivables.

36:56And the risk factors admit that they've sought quantities of compute from CSPs in excess of what those partners were willing to provide. So being on all three major clouds is the biggest benefit to the go-to-market. Now they've started to vertically integrate and build out their own compute on top of that. The total purchase commitments and lease agreements are blank here. And I'd hate to be the lease accounting person at that company. T's and P's. Potential red flag. So the risk section is, I'm not going to lie, it is robust. My dog's even growling it's so robust. So here are the five that I would pay attention to.

37:28I'll try to do it in plain English for you. The first one, everything is connected to everything. So quote, many of the risks we face are interconnected and may compound or trigger cascading adverse effects across our business, which could amplify their overall impact. That is scary. So it's like a vortex of hell. Their own example, a bad copyright ruling forces them to retrain or limit their models, which hurts revenue, which makes it harder to pay for compute, which could trip covenants on their credit facility and trigger cross-defaults elsewhere, like the broader U.S. economy. Have you seen The Big Short?

37:58That was a good movie. Quote, Because these and other risks may interact in complex and unpredictable ways, we may not be able to foresee or effectively manage all potential consequences. Two, the chips come from Taiwan and the power comes from regulators. So it's pretty nuts to me that an island about the size of Maryland and Delaware combined is like the linchpin for the AI economic engine. By most estimates, TSMC makes about 90 % of the world's most advanced chips in the filing lists, the exposure and how earthquakes and typhoons and tensions with China and all that stuff could really impact the supply chain.

38:34And they say efforts to build capacity elsewhere remain in early stages. On power, by the way, they need deals with utilities and grid operators plus sign off from public utility commissions. Not in my backyard. Three, the bills show up before the buildings do. So, quote, large upfront investments and long-term payment commitments must be made months before facilities become fully operational. It's like building a big ass house. Most of the big compute deals are non-cancellable, by the way. So if you're in for a penny, you're in for a pound. And they left that part blank. Thank you. Number four, copyright is still alive.

39:09Why? There was this case. Anthropics spent millions of dollars buying print books. They ripped off the bindings and cut out the pages and scanned them one by one and threw the originals away. So picture like a first edition of The Great Gatsby going under the paper cutter. That would truly make my second grade librarian, Mrs. Walters, weep. The judge actually said the scanning was totally cool, but wasteful. This$1.5 billion they paid to settle with BARTs in 2025 was actually for the pirated books they downloaded virtually. The bigger risk now is fair use not holding up for AI training in other cases, and that's spiraling down and forcing them to change their models.

39:47Number five, your vote does not really count for much. So the Long-Term Benefit Trust, the LTBT, elects a majority of the board and the founders hold 50.1 % of the vote. The filing says the structure may effectively eliminate the ability of holders of our Class A common stock to influence important decisions, including a change in control. So nobody is pulling a Kendall Roy here and overthrowing the board. Some honorable mentions. Quote, we have a limited history selling consumption-based access, says the largest usage-based company in the history of the world. Free tiers might not convert and users may choose to use our free offerings indefinitely.

40:23Like my friend Kyle Poyer. Give him shit if you talk to him. He doesn't pay for any chat GPT or Anthropic. He's cheap. Next, they flag unfavorable media coverage as a risk. So, I mean, give me a TBPN deal and I'll make CFOs fucking love you. Good news, though. They get the green light on the following. no material weaknesses in internal controls, no going concern language, nearly$132 billion of cash in short-term investments, and no single customer over 10 % of their revenue. Cap table. There are five classes of common stock. Two of them are a single share. I don't think I've ever seen that. Class A, that's what you're going to buy in the IPO if you want it.

41:00One vote per share that doesn't really count. Class N, same economics as class A, but no vote, held mostly by strategic partners like Amazon and Google. That makes sense. Converts to Class A when sold to someone unaffiliated. This is also the stock that funds the 5 % public benefit fund. Class T, one share held by the LTBT Alexa majority of the board. They did a one for a 1000 reverse split to get it down to exactly one share. Class F, one share held by an LLC owned by the seven co-founders. That's someone that carries 50.1 % of the vote. That doesn't seem to be permanent. Voting power passes to common stockholders, ones fewer than three co-founders remain.

41:40Classy, that one's for the employees who are not founders. Who owns this? Interestingly, no institutional investors show up on the plus 5 % table. Amazon and Google mostly hold non-voting class N, so they don't have to be listed. The seven co-founders each hold about 240 million class A shares. All seven are listed as selling something. If you are close to securing the bag, make sure you secure the bag. Ownership percentages are blank along with how much they'll sell. The co-founders have each pledged 80 % of their equity to philanthropy. So how the execs get paid. Dario CEO, 18 million total in 2025.

42:15That was 654K in salary, 9.3 million of RSUs and 4 million of options, 2.7 million of personal security. He definitely needs that. 97 ,000 of personal private air travel required by a security program. That's like one flight to Florida. I'm back 1.15 million from the spread on shares. He sold back in June, 2025 tender offer. Daniela Amodi president, 16.4 million with the same salary and equity is Dario. Krishna Rao, their CFO 720 K total in 2025 with no new equity that year. His 2026 base salary is 1.25. So he's going to make more in cash than Dario and Daniela's 915 ,000. He's also getting company paid security right now.

42:59The company paid 2.5 million each to cover Dario and Daniela's HSR antitrust filing fees, which they owed because of how much stock they hold. The severance in here, if they get fired, is 24 months of salary for Dario and Daniela and 12 months for Rao, plus a year of accelerated investing, which is actually worth way more. The board is seven directors. The trusts will elect four of them and the co-founders elect the other three. Reed Hastings of Netflix fame, Chris Liddell, who is not Chuck Liddell, the UFC fighter, someone named Voss, and someone named Yasmin of Spark Capital alongside Dario and Daniela.

43:36Valuation. The price range is blank, but the reporting has bankers pitching about$2 trillion, roughly double the$965 billion series H from five months ago. They're also sitting on about $132 billion of cash in short-term investments before a dollar of proceeds come in, so they're doing okay enterprise value is closer to 1.87 trillion after factoring for that so what does 2 trillion buy really depends on what number you pick pick a card any card so trailing 12 month gap revenue through june so this is the most intellectually honest version so 19.7 billion that will get you to 101x multiple wow if we go off our backed into ttm revenue through september with some assumptions that's 35 billion that's only 57x revenue only q2 revenue times four that's 46 billion or 43x.

44:24The 65 billion run rate reported for July, that brings it down to 31x. The 100 to 120 billion year end run rate investors expect per the Financial Times gets you to a paltry 17x. I mean, the multiples anywhere from like 17 to 100, depending on whether you believe the run rate or the gap. And if you strip out the cloud providers cut from the gross booked revenue, every one of those multiples goes up by about a fifth. Remember, because they say it's gross. For context, here's what recent names landed in our breakdown this year. So SpaceX asked for 60 to 70x forward revenue. Cerebris came out around 50x when it was said and done.

45:00N-Scale is asking about 90x in its run rate at the time. Palantir trades around 36x today, but earlier this year they traded 100. NVIDIA is around 18x at the time of our SpaceX piece when we published that. Miscellaneous stuff a note. First, they listed their Threads account just in case you need to contact them. That was in the contact section. Number two, they switched auditors year before filing. They dismissed Grant Thornton and engaged EY in October 2025. Sponsor the pod. They cut the vesting cliff in half. In January 2026, they changed the one-year cliff to six months on all stock options they hadn't hit yet, which is a nice recruiting move.

45:36Their boomerang rate is super high. So of the employees who left, they wanted to keep. 19 % came back. You know, apologies and shout out to Societe Generale. Is it Societe Generale? They were listed in the fourth row of the 22 banks. This bank caught a stray by me in my SpaceX breakdown, but I gotta say, you punch above your weight class. So the bull case is basically the efficiency section that I went through. Like this business is growing 10X a year with 900 % net revenue retention. If you price it like based on their run rate, you're only looking at 20 to 30X multiple, which is the last time a CrowdStrike and Cloudflare get on forward revenue with a fraction of their growth.

46:13On trailing gap revenue though, you're paying about 95X. That's a lot more than Palantir for a company that buys its compute from Amazon, Google, and Microsoft who sell competing models on the same clouds that resell Claude. And it all hinges on my number one question, which is whether the use cases keep showing up as fast as the price per task keeps falling. And finally, never let your company be worth a dollar more than it is actually worth. They said it, not me. Remember, none of this is investment advice. I wrote it at my kitchen table full of pancakes with my dog, Walter, waiting for a bite that he never got.

46:48Do your own homework. This is for information and entertainment purposes only. Wishing you a short stack, a long lockup, and the real maple syrup. Peace. Run the Numbers is a mostly media production. Yelling an intro by Fat Joe. Artwork by Meg D 'Alessandro. Nothing said on this podcast is intended to be business or investment advice. It's the sole opinion of me, a guy who feeds his dog way too much ice cream and has a history of net operating losses. LOL. If you like this podcast, hit subscribe and give us five stars. It will take like two seconds and our algorithm overlords love it. Drink water, call your mom and have a great day.

47:24Peace.