Anthropic just told the market what it thinks it’s worth. The number everyone will repeat is $1.404 trillion. It’s the wrong number to fixate on.
In an updated S-1 filing on Tuesday, the Claude-maker said it would offer 825 million Class A shares at $70 to $78 apiece, raising between $57.75 billion and $64.35 billion. At the top of that range, on a fully diluted basis, the company is worth about $1.4 trillion. It starts trading on the Nasdaq under the ticker “ANT” as early as November 19.
That’s a monster number. It’s also not the number that tells you what’s actually happening.
The $2 trillion whisper that died
In May, Anthropic was valued at $965 billion in a fundraise. A couple of months ago, the whisper number around its IPO was $2 trillion. This week, the company priced its range at a fully diluted valuation of $1.404 trillion — and on a non-diluted basis, excluding options and restricted stock, closer to $1 trillion.
CNBC reported last week that some Anthropic investors were already bracing for a lower valuation, with one calling even $1.5 trillion a “stretch.” Fortune put it plainly: the range is well short of the $2 trillion that some news outlets had reported just weeks ago.
Here’s how I read that. We’ve spent two years treating AI valuations as a momentum trade — the number goes up because the story gets bigger. This is the first time the story met a price. A trillion-dollar company came to market and the market said, “Show me the revenue.” Even the hottest asset in tech has a ceiling, and the ceiling is lower than the hype.
We’ve seen the same dynamic before. When China’s AI bubble burst, the $480 billion meltdown revealed how much of the “AI economy” was really just capital spending with a story attached. Anthropic didn’t have a meltdown. It had a repricing — which is what a mature market does.
To Anthropic’s credit, $1.4 trillion would still be the second-highest valuation in history for a newly public company, behind only SpaceX’s June debut. That’s not a failure. It’s a number with a ceiling.
The $26 billion that isn’t in the headline
Buried in the same filing is a deal that tells you more about the AI industry than the valuation does. Anthropic has agreed to privately sell $26 billion of non-voting Class N stock at the IPO price to seven strategic partners: AMD, Broadcom, Google, Micron, Microsoft, Nvidia, and TSMC. Google and Nvidia are each putting in $10 billion.
Read that list again. Those are the companies that make the chips, the memory, and the networking that Anthropic runs on. The suppliers are now the buyers.
That’s not a coincidence, and it’s not charity. When the company that sells you the GPUs also becomes a shareholder in the company buying the GPUs, you’ve locked in demand years in advance. It’s vertical integration with extra steps — the entire AI stack, from silicon to model, now has a financial stake in the model-maker’s survival. I covered Anthropic’s $3.4 billion turbine bet yesterday; this is the same impulse, scaled up. The company is securing its supply chain by making its suppliers its owners.
It also means the IPO isn’t really a liquidity event for the ecosystem. It’s a financing round where the lenders happen to be the vendors.
The founders cash out, but they keep the keys
All seven co-founders plan to sell exactly 3,378,378 shares each in the IPO. At the $74 midpoint, that’s just under $250 million apiece. At the low end, about $236 million; at the high end, just under $264 million.
After the sale, each will own roughly 3.2% of the Class A shares, or about 22% collectively. That sounds like a lot of dilution. It isn’t, because of how Anthropic structured the deal.
As a public benefit corporation, the founders hold Class F shares carrying supervoting rights. Collectively they’ll control 51.1% of the votes. The Long-Term Benefit Trust will elect a board majority. Class A buyers — the public — get dividends and a price chart, not influence.
Think about what that means. The company that has built its brand on AI safety and responsible development is going public in a structure that insulates it from public-market pressure. The founders can take a quarter-billion dollars each and still not answer to anyone but the trust. If you’ve been following Anthropic’s decision to let outside auditors in, this is the other side of that coin: openness on safety, control on governance.
Board member Reed Hastings has indicated he wants to buy $300 million worth of shares through his donor-advised fund. That’s a vote of confidence — and a reminder that the people closest to the deal see the range as a discount.
The economics nobody’s pricing
Now the part that should make you pause. Anthropic’s run-rate revenue hit $81 billion by the end of October — about $10 billion more than at the end of September. Third-quarter revenue was $17 billion. Nine-month revenue was $33.2 billion.
The nine-month net loss was $104 billion.
Let that sit. The company is losing more in nine months than it will raise in the entire IPO. And the reason is sitting in the same filing: compute and infrastructure commitments of $900.9 billion. Anthropic has nearly $132 billion in cash and short-term investments, and it has promised close to a trillion dollars to the people who build its data centers.
This is the part of the AI trade that doesn’t fit in a bull case. The frontier labs aren’t software companies with a margin problem. They’re infrastructure companies with a revenue problem. Every dollar of GPU time has to be pre-ordered years out, and the bill comes due whether or not the models sell. I wrote about the $3.2 trillion debt nobody talks about last week; Anthropic’s own filing is the case study.
The IPO isn’t an exit. It’s a bridge loan for the next build-out.
What this means for the rest of us
If you’re a developer or an enterprise buyer, the lesson is about dependency. The same seven companies that now own a piece of Anthropic are the ones selling you the hardware to run it. Pricing power doesn’t accrue to the model layer — it accrues to the stack underneath. That’s worth remembering the next time someone tells you the API price is going to keep falling.
If you’re watching from the Philippines, the takeaway is different but related. We’re building on top of models that are being financed by the people who sell the chips. The cost of the tools we use, and the terms we get them on, are being set in a room we’re not in. The practical move is the one I keep coming back to: own your data, keep your options open, and don’t build your whole operation on someone else’s runway. OpenAI’s revenue gap showed the same pattern — a frontier lab whose spending had outrun its sales long before the IPO window opened.
Anthropic’s listing will be the biggest tech debut in years. But the number that matters isn’t $1.4 trillion. It’s the $26 billion in supplier lock-in and the $900.9 billion in compute commitments. Those are the numbers that tell you where the AI industry is actually going.