Silicon Valley wants you to invest in miracles

Anthropic's S-1 runs 335 pages, and 94 of them are risk factors. Somewhere in that stack the company tells would-be shareholders that the technology it sells could pose catastrophic or existential risks to humanity, and that it may never reach sustained profitability.

That is not normal investor language. When a frontier lab goes public, ordinary retirement accounts and index funds can end up holding a piece of that bet. How the company frames the risk now shapes what investors, regulators, and employees expect it to do later.

We'll say the uncomfortable part first: we find a lot to respect in this filing. The founders call their approach an "anti-sell," warning buyers away from expectations the company cannot meet. Putting the risk of models behaving in ways nobody intended on the record, in a legal document, is better than burying it in a keynote. If useful AI is going to be built at scale, we would rather the builders say the dangerous part out loud.

But candor is not the same as a plan. A prospectus that lists every conceivable harm can also make the important ones harder to find. The University of Florida's Jay Ritter makes exactly that point: bloat makes it harder to spot what matters. Apple's 1980 prospectus had 47 pages and zero risk factors. Today's AI filings read like manifestos with an appendix of fears attached.

What I'm watching most closely is the gap between the poetry and the numbers. The filing leaves out third-quarter results. It shows strong second-quarter growth during what The Verge calls the height of tokenmaxxing, along with hundreds of billions of dollars in compute obligations and a plan to build its own data centers. Narrower losses are encouraging. But the obligations are huge, and the revenue trend has only been shown for a short window.

The public benefit corporation structure is the strongest argument Anthropic has. It legally lets the company weigh its mission alongside returns. Still, it's hard to square a stated principle of underpromising and overdelivering with a letter that promises an age of miracles. If this becomes the template, every AI IPO will be sold as destiny. Investors who are told to just believe may well end up holding the bag when a quarter disappoints.

Our read: this is the most honest risk document a major AI company has produced, attached to one of the most aggressive financial bets in tech. Both things are true, and investors should price both.

Anthropic filed the S-1 on Monday and will go public as a public benefit corporation, a structure that requires it to balance a social mission with shareholder interests. The comparisons are SpaceX's June IPO filing, Facebook's 200-page S-1 with its "Hacker Way" letter, and WeWork's 2019 filing, which was mocked for vowing to elevate the world's consciousness. CEO Dario Amodei has previously put the chance that further AI development leads to human extinction at up to 25 percent.