Anthropic IPO: Did It Really Lose $80 Billion? The S-1 Explained

0:00 $80 billion gone in 6 months. That is the headline and it is wrong. The truth hiding inside this document is crazier than the headline. And if you have money in the stock market, it may soon concern you directly. On October 26th, Anthropic, the company behind the AI assistant Claude, filed to go public on the NASDAQ under the ticker ENT. This could be the biggest stock market debut in history, bigger than SpaceX. And here's the part nobody tells you. When a company this size lists, it usually flows into the big index funds. So, if your savings or your pension track the market, you may soon own a piece of this company, whether you chose it or not. I build AI and data systems for financial companies for a living.

0:47 So, I did what almost nobody does. I read the actual filing, all of it. In the next 10 minutes, three numbers that look impossible. The risks Anthropic admits in its own words and the one number that fooled almost every headline on the planet. That last one will change how you read financial news forever. Let's go. Number one, growth that breaks the charts. 3 years ago in 2023, Anthropic's revenue, the money coming in from customers, was $41 million. The next year, 386 million. Last year, 4.6 billion. And in the first 6 months of this year alone, 16.3 billion. Let me repeat the craziest part.

1:34 The second quarter of this year, just 3 months, brought in 11 12 billion. That single quarter was more than two times the entire previous year. Where does the money come from? Mostly from companies. The filing says more than 11,900 customers pay at least $100,000 per year. And now the number that genuinely shocked me as a finance person, net revenue retention of around 900%. In simple words, take last year's customers and check how much the same customers pay this year. For a normal very good software company, that number is around 120%. The same customers pay 20% more and investors celebrate. Anthropic's number is 900%.

2:20 The same customers pay nine times more than the year before. I have never seen that in a filing. One more revenue per employee is around $6.6 million. Several times higher than even the best software companies. But, and this is a big butt, growth like this has a price tag. And that price tag is the scariest number in the whole document. Number two, the $637 billion bill. Training and running AI models needs enormous computing power. The industry calls it compute. Thousands of powerful chips running in data centers day and night. Antropics filing list signed compute contracts of roughly $637 billion with Google, Amazon, Microsoft, Broadcom, and even SpaceX's AI business.

3:13 That is more than the yearly government budget of most countries on Earth. So, is this company just burning money? Here is where it gets interesting. Look at the trend, not the total. In the first half of last year, Anthropic spent about $2 on compute for every $1 of revenue. Spend two, earn one. A terrible business. But in the second quarter of this year, that number fell to 48. Spend 48 cents, earn $1. Why so fast? Two reasons. Revenue is growing much faster than compute costs. In the second quarter, revenue grew about 14 times year-over-year, while compute costs grew about four times. And each new model does more work for less money.

3:59 The filing says the newest clawed model cut customer cost per task by about 40% compared to the previous one. The result in the second quarter, for the first time in its history, Anthropic made an operating profit, $974 million. Operating profit means revenue minus the real day-to-day costs of running the business. So, the machine is starting to pay for itself. But before you get excited, the filing also has a risk section. And the three things Anthropic admits in there would scare most investors if they ever read them. You are about to. Now, the risks in Anthropic's own words. Risk one, customer concentration. About 20% of revenue comes from just two customers. The filing does not name them, but [music] if one of them leaves, that is billions gone overnight.

4:49 Risk two, and this one is wild. Anthropic's biggest suppliers are also its biggest competitors. The compute comes from Amazon, Google, and Microsoft, the same companies building their own AI models. Anthropic writes that it struggles to secure enough compute [music] because these companies increasingly treat it as a competitor, not a customer. The filing even warns the compute could be quote curtailed, repriced, or terminated. Imagine your business depends on a supplier who wants to beat you in the market. Risk three, control. When you buy ENT shares, you get almost no voting power. The seven co-founders keep [music] 50.1% of all votes through one special share class.

5:35 Whatever shareholders think, the founders decide. This is common in tech IPOs, but you should know it before you buy. And one more unusual sentence. Anthropic warns that a catastrophic AI incident could cost more than its insurance covers. Read that again. The company itself puts that risk on paper. So amazing growth, improving economics, serious risks. But none of this, none of it is the number everyone got wrong. That number is the $80 billion loss from the headlines. And this is where it gets really interesting. The headlines say Anthropic lost $80.7 billion in 6 months. Sounds like a catastrophe, right? Now, open the filing and look closer.

6:21 About 72 billion of that loss is not real money leaving the company. It is a paper loss, an accounting entry. Let me explain it in one minute. Before the IPO, Anthropic borrowed money using convertible notes. A convertible note is a loan that later turns into shares instead of being paid back in cash. Accounting rules say the company must report these notes at their fair value, what they are worth right now. And here is the twist. When the company's value goes up, those notes become worth more because the lenders will receive more valuable shares and the company must book that increase as a loss. Read that again. The better Anthropic performs, the bigger this paper loss looks. The $80 billion catastrophe from the headlines is largely a side effect of the company becoming extremely valuable extremely fast.

7:14 So what is the real number? The operating loss, real revenue minus real costs, was $693 million for the 6 months. And remember, the second quarter inside that period was already profitable. Is 693 million still a loss? Yes. Is it the same story as 80 billion lost? Not even close. This is why I always say don't read the headline, read the filing. The headline is written to be clicked. The filing is written for regulators and it has to tell the truth. So let's put it together. The bull case, the fastest revenue growth ever seen at this scale. 900% retention, unit costs falling, and the first profitable quarter. The bare case, a $637 billion compute bill, two customers equal to 20% of revenue, suppliers who are competitors, and a price tag that may pass $1 trillion for a company that has been profitable for exactly one quarter.

8:17 I am not a licensed financial adviser, and this is not investment advice. I am showing you how to read the numbers, not telling you what to do with your money. But I want to know what you think. Would you buy a trillion dollar company with one profitable quarter? Yes or no and why? Write it in the comments. I read every single one. If you work in finance and want to actually [music] understand AI, not the hype, the real thing, that is exactly what this channel is about. Subscribe and I will see you on listing day when [music] we find out the final price. See you in the next one.