Anthropic’s IPO Prospectus Reveals What’s Under the Hood

Anthropic has publicly filed its S-1 and plans to list on Nasdaq under the ticker ANT. The Wall Street Journal has reported the IPO could come later in November. The filing does not say how much the company wants to raise or at what valuation, though The New York Times reported in August it was looking at more than $100 billion at roughly a $2 trillion valuation.

This is the first time one of the frontier labs has opened its books this way. OpenAI filed confidentially over the summer, then said in September it no longer intended to go public this year. So for now, Anthropic's prospectus is the clearest public look anyone has at what it costs to build frontier AI, who is paying for it, and who stays in control.

Start with the business, because it is genuinely strong. TechCrunch and The Verge report $16.3 billion in revenue for the first half of 2026, against $1.2 billion in the same stretch last year. The enterprise base is the real story: more than 11,900 customers spending at least $100,000 a year, about 1,900 spending $1 million or more, and roughly 250 spending $10 million or more. That is not hype money. That is companies putting Claude into work they actually pay for, and we think that's the strongest evidence yet that useful AI has a durable market.

Now the costs. The reported first-half net loss is $80.7 billion, which both outlets say is driven largely by noncash accounting charges tied to convertible notes. The operating loss tells a cleaner story: $693 million, down from $2.1 billion a year earlier. Then there's compute. The Verge details commitments that include about $3.1 billion a month to SpaceX for Nvidia GPU access once capacity is delivered, at least $114.5 billion with Google through 2033, $110 billion with Amazon over ten years and $44.9 billion with Microsoft. Those are obligations that assume demand keeps climbing. If it stalls, they don't go away.

The governance is where I'd slow down. The seven co-founders, through a Founder LLC, will control 50.1% of shareholder voting power, and the Long-Term Benefit Trust can appoint most of the board. Anthropic frames this as protecting its safety mission. That may be sincere, and there's a reasonable argument that a safety-focused lab shouldn't be steerable by quarterly pressure. But it also means public shareholders get very little say over a company that warns, in its own filing, that advanced AI could pose catastrophic or existential risks. Accountability is being placed on trust in a small group and a trust structure, not on owners.

The risk section deserves credit for candor. It names misuse for cyberattacks and novel bioweapons, authoritarian surveillance, the possibility of losing control of its own models, copyright fights, AI-specific taxes, even model sentience. It says no single insurer, and maybe not the insurance industry as a whole, could underwrite a catastrophic AI incident. It also flags the administration's supply-chain-risk designation and related lawsuits as a threat to government sales. That last one is a live business risk, not a thought experiment.

Our read: this is a serious company making a serious product, asking the public market to fund an enormous bet while keeping the steering wheel. The language asking investors not to think of the stock in purely transactional terms is unusual for a prospectus. If this works, it becomes a template for how mission-driven labs raise capital at scale. If it doesn't, the people least able to influence the outcome will be the ones holding the shares.

Anthropic was founded in 2021 by seven former OpenAI employees and has built its brand around safety. Its CEO, Dario Amodei, has repeatedly warned that AI could eliminate large numbers of jobs and act against human interests. The filing also discloses that he received nearly $18 million in 2025 compensation, mostly in stock and option awards, and that president Daniela Amodei received $16.4 million.