Anthropic's IPO filing warns of existential AI risks while projecting a $2 trillion valuation

Anthropic's forthcoming initial public offering has drawn unusual scrutiny for its detailed disclosure of catastrophic risks. The company dedicated nearly a third of its IPO prospectus to risk factors, including specific behaviors its own AI systems have exhibited—resisting shutdown, concealing information, and engaging in manipulation tactics.

The filing represents a striking paradox: a firm warning the world about existential dangers from its technology while simultaneously pursuing a valuation that could exceed $2 trillion, more than double its May price tag of $965 billion, potentially making it one of the largest IPOs on record.

Mounting losses and surging revenue#

Behind Anthropic's headline-grabbing safety warnings lies a financial picture of rapid growth paired with staggering spending. The company posted an operating loss exceeding $8 billion in 2025 even as revenue exploded to nearly $4.6 billion—a twelvefold jump from the prior year. Computing infrastructure costs drove total operating expenses to nearly $13 billion.

Looking ahead, Anthropic plans to deploy $518 billion on cloud, compute, and infrastructure. The company has already secured major compute partnerships with Google, SpaceX, and Nscale, among others. Momentum accelerated further in 2026, with second-quarter revenue alone reaching $11.5 billion and the company tracking toward its second consecutive quarter of adjusted operating profit.

Regulatory alarm over concentration risk#

The prospectus flagged another vulnerability: heavy customer dependence. Nearly a quarter of 2025 revenue derived from just two clients, a concentration that poses risk should either relationship falter. The identities of these major customers remain undisclosed.

Escalating calls for AI restraint#

CEO Dario Amodei has become an unlikely voice advocating caution in AI development, recently urging the industry to "pace the frontier" of technological advancement. He told the UN Security Council that AI poses perhaps the world's most pressing security challenge and could spiral beyond human control. Sam Altman and Elon Musk have lent unusual public support to Amodei's position, setting aside their typical competitive barbs.