Anthropic goes public in the biggest IPO ever
The maker of Claude raised more money in its IPO than any company before it. If you build on an AI model, its maker now answers to the stock market every quarter too.
What happened
Anthropic, the company behind the AI model Claude, has been listed on the Nasdaq since Thursday 19 November. At about $93.5bn, it is the biggest IPO in history.
On Wednesday 18 November Anthropic priced its shares at $83, above the $70 to $78 range it had marketed to investors on its roadshow. Demand was strong: according to people familiar with the deal, orders reached about six times the shares on offer midway through that week, and close to seven times by Sunday. Anthropic therefore sold 980 million shares instead of the planned 825 million, and the banks used their option on 147 million more. On top of that, AMD, Broadcom, Google, Micron, Microsoft, NVIDIA and TSMC Partners bought $26bn of non-voting shares at the same price.
The first trade only came more than two hours after the bell, at around 16:50 UTC: $111.50, 34% above the offer price. The shares then slipped back to close at $97.17, still 17% up, on a day when the S&P 500, the Nasdaq and the Dow Jones all fell more than 2%. Before Friday’s open, according to StockTwits, the stock gave back roughly another 6 to 7%, amid a wider sell-off in AI and chip stocks.
At the Nasdaq, chief executive Dario Amodei promised that Anthropic would keep training and releasing AI at a safe pace, and would not bend to pressure from business or politics. The seven founders sold about $1.95bn of shares between them. Each has pledged to give 80% of their shares to charity. According to TechCrunch, the prospectus calls these “voluntary, revocable” expressions of intent that are not legally binding.
We will not bend to either business nor political pressure.
Dario Amodei at the Nasdaq on 19 November 2026, as reported by NBC News
The background
Why the figures differ
The coverage gave proceeds from $81bn to nearly $120bn and valuations from $1.5tn to $2tn. They are all right, but they count different things. The $81.3bn reported by Al Jazeera, TechCrunch and others is the base deal: 980 million shares at $83, of which 23.6 million came from existing shareholders. With the banks’ extra shares, the IPO came to $93.5bn according to Anthropic itself, of which $91.6bn went to the company. Add the $26bn from the seven tech companies and Anthropic raised $119.5bn.
For the valuation, the moment is what differs. At $83, Anthropic is worth about $1.5tn according to Fortune and the Financial Times, counting all shares and options. At the first trade of $111.50 NBC News put it at about $2tn, and at the close at $1.76tn. Our arithmetic: all three assume roughly 18 billion shares; only the price differs. SpaceX raised $75bn in June, over $85bn with its extra shares, and at $1.77tn at its offer price it remains the most highly valued IPO.
Who is in charge
Anthropic is a public benefit corporation: a US company form in which the board must weigh profit against a public purpose, here the responsible development of advanced AI for the long-term benefit of humanity. Since 2023 there has also been the Long-Term Benefit Trust, a group of trustees with no financial stake in the company. Through its own class of shares, the trust elects and removes a majority of the board. The founders hold 50.1% of the vote on other matters, such as electing the directors the trust does not choose.
So a new shareholder has little say, and the prospectus says as much: according to the Financial Times it asks buyers not to think of their purchase in purely transactional terms. Amazon and Google together own 28% of Anthropic.
Fast growth, a big bill
In 2023 Anthropic had about $41m in revenue. In the first nine months of 2026 it had $33.2bn, thirteen times as much as a year earlier. In the prospectus Anthropic itself says it does not expect to keep up that growth. Against it stand commitments for compute and data centres of more than $1tn according to TechCrunch, $70.7bn of it in the next twelve months. After the IPO the company has almost $230bn in working capital.
Our analysis
The maker of your AI model now reports every quarter
For the market, this is a record story. For an organisation building on Claude or any other model, the news lies elsewhere. A listed model maker shows figures every quarter, and those figures weigh on choices about prices, models and terms. Anthropic’s board is set up so that investors do not elect it, but it will feel the share price all the same. Our expectation: prices and the line-up of models will move more visibly with the quarterly figures, as at any listed software company.
What the coverage underplays: going public also makes Anthropic more transparent than ever. The prospectus shows what customers rarely get to see, such as the cost of compute and the ties to a few large partners. And according to The Batch it records that in June the US Department of Commerce imposed worldwide export restrictions on Fable 5 and Mythos 5, two configurations of what was then Anthropic’s most capable model. Anthropic switched them off for every customer until the restrictions were lifted. Every customer felt that, wherever they were.
So the lesson is not a choice for or against one maker. Make sure your AI work can move, and know where your data is processed.
What stands out
- A share, but little say The trust elects a majority of the board, and the founders hold 50.1% of the vote on many other matters. New shareholders mainly buy a chance of a return.
- Up on a day the market fell The three main US indices lost more than 2% on Thursday. Anthropic still closed 17% above its offer price.
- Suppliers became shareholders Companies that supply Anthropic with chips and compute, such as Google and Broadcom, bought in for $26bn, without voting rights.
- A pledge to charity Each founder pledged to give away 80% of their shares. The prospectus calls this voluntary and not legally binding.
What it means for you
If you build on an AI model, part of your work depends on the choices of one maker, and that maker now has the stock market to answer to as well.
- Keep your AI work portable Keep your own copy of your instructions, test questions and good example answers. Then you can test another model in days rather than months.
- Know where your data is processed With which maker, in which country and on what terms. For the most sensitive work, an open model on hardware you control can be the foundation.
- Keep an eye on prices and terms They can change when the quarterly figures call for it. If you keep track, you will not be caught out.
Where we come in · Advisory
We make your AI work less dependent on one model
A Flaire adviser works out with you which AI applications you use, which model they run on and what happens if that model gets more expensive or is briefly unavailable. And for work that must not leave the building, we install an AI server in your own office: a Mac Studio, cut off from the internet, running the latest open models.
Do you know which AI model your work runs on, and how quickly you could switch?
Tell us which AI applications you use. We’ll think it through with you and always reply within one working day.