Anthropic IPO filing warns of 'catastrophic' AI risk, $518 billion spend

Anthropic's filing reportedly pairs one of the largest planned infrastructure commitments ever disclosed in an IPO — $518 billion for cloud, computing, and infrastructure obligations — with a $42 billion net loss for 2025, even as revenue rose 12-fold to nearly $4.6 billion. The Financial Times reports that nearly a quarter of 2025 revenue came from just two clients, a concentration that makes the growth story more fragile than the headline revenue figure suggests.

The governance plan is unusual even by tech-founder standards. A 'Founder LLC' of Dario Amodei and six co-founders would retain 50.1 percent voting power, and the company is structured as a Public Benefit Corporation under Delaware law with the stated aim of keeping Anthropic focused on responsible AI. That structure may appeal to some investors but could raise questions about accountability to public shareholders.

What stands out most is the self-incriminating risk section: 80 pages of the 261-page filing devoted to how Anthropic's own models could cause catastrophic harm. Whether that candor strengthens or weakens the IPO likely depends on whether investors view it as responsible disclosure or as a material threat to the business model itself.