Math first. Anthropic priced its IPO at $83.00 per share, across 980 million Class A shares. That’s roughly $81 billion at the base deal size, before the underwriters’ option for another 147 million shares. If that figure holds up, it is one of the largest public listings ever, and it belongs to a company that sold its first product only a few years ago.
Of the shares offered, 956,351,354 come from Anthropic itself. The other 23,648,646 are sold by existing stockholders, and Anthropic gets nothing from those. Trading starts November 19 on the Nasdaq Global Select Market under the symbol ANT, with closing expected November 20. Goldman Sachs and Morgan Stanley are the names on the prospectus.
Why does this matter beyond the finance pages? Because training frontier models costs staggering amounts of money, and private rounds have been the main way to pay for it. Anthropic leaned on Amazon and Google for years. OpenAI has relied on Microsoft and giant private raises. A public listing gives Anthropic a much bigger and more durable pool of capital, and it can use its stock to pay for compute, acquisitions and talent.
But there’s a cost. Anthropic is a public benefit corporation, hence the PBC in its name, and it has built its brand on safety commitments. Public investors want quarterly growth. Those two things can pull against each other, and the first earnings calls will show which one wins when they do. Expect analysts to press hard on margins, compute spending and revenue concentration in enterprise and API customers.
For developers and founders, the practical effect is transparency. Until now, Anthropic’s financials were mostly rumor and leaks. Filings will now show real numbers on revenue, losses and infrastructure commitments. You’ll be able to judge whether Claude pricing is sustainable, and whether a platform you build on has the runway to last.
The competitive pressure is obvious too. OpenAI, Google DeepMind and xAI all need huge capital, and a successful ANT debut raises the odds that others follow. A weak debut would chill that. Either way, AI labs are now being priced like infrastructure companies, not research projects. Whether $83 looks smart or stretched will be settled by the market within weeks, but the shift has already happened.



