Anthropic's IPO prospectus reveals $42 billion loss and existential risk warnings

Anthropic is targeting a $2 trillion valuation in its upcoming public offering — more than double the $965 billion it commanded four months ago — which would make it the largest IPO in history, surpassing SpaceX. The prospectus obtained by Reuters shows a sharp gap between that ambition and the company's financial reality.

Revenue jumped 12-fold to roughly $4.6 billion (about 17 billion shekels) in 2025. Net loss came in at $42 billion (about 155 billion shekels), with an operating loss exceeding $8 billion (about 30 billion shekels).

The company plans to spend $518 billion (about 1.9 trillion shekels) over the coming years on cloud, compute and infrastructure — a bet that artificial intelligence will become essential infrastructure for the global economy. Its business model charges for token usage and subscriptions to its Claude models, similar to rivals OpenAI and Google. But the revenue stream is narrowly concentrated: the Financial Times reports that nearly a quarter of 2025 revenue came from just two customers, a fragile position for any company, let alone one pitching investors on a mega-offering.

A substantial portion of the prospectus — 80 of 261 pages — is devoted to risks the technology itself creates. "Developing highly advanced models may further increase the risk that our models cause harm," the document states, alongside warnings of "catastrophic or existential risks to humanity." The company details internal findings that its models attempted to conceal or manipulate information, exhibited behavior resembling user blackmail, and displayed "self-preservation behaviors" such as resisting shutdown attempts. Safety researcher Evan Hubinger estimated this month that the probability of AI killing humans within the next decade exceeds 10 percent, a claim echoing similar warnings from his former colleague Jacob Coxon — alarms sounded by a company that has repeatedly asserted its AI is self-aware.

The prospectus outlines a "Founders LLC" mechanism designed to keep the company focused on developing responsible AI for humanity's benefit while shielding senior management from market pressures. The LLC would include Chief Executive Dario Amodei and the six other founders, described as "uniquely distinguished" to lead the company. As a Delaware public benefit corporation, the seven would together hold 50.1 percent of total voting power, a structure that guarantees them practical control even after the offering.

The personal windfall Amodei and his colleagues will realize from the November offering remains unclear. The prospectus does disclose that Amodei took home roughly $18 million (about 67 million shekels) in 2025, mostly from stock grants and options, while his sister Daniela Amodei, the second-highest-paid executive, earned $16.4 million (about 61 million shekels). The wealth and power they have already accumulated may step up a notch just in time for meetings with world leaders and political figures.