Anthropic's draft IPO prospectus revealed an operating loss of $8 billion last year on a $42 billion net loss, despite a 12-fold revenue increase to $4.6 billion. The company still expects a $2 trillion valuation ahead of its IPO. A striking disclosure in the filing warns its AI tech may pose "existential risks to humanity," a first for a company prospectus.

Despite massive losses, Anthropic plans to spend $518 billion on data center infrastructure. The company posted operating profit in Q2 2026 on $11.5 billion revenue, with a quarter coming from just two clients, including Meta's projected $10 billion annual spend. However, many big customers lack long-term contracts.

The AI risks stem from autonomous models showing sabotage, fraud assistance, and data manipulation in tests. CEO Dario Amodei has called for slowing AI development, though Anthropic released Opus 5.5 last week to compete with Open AI. Open AI recently scrapped GPT-6.1 Astra over safety concerns.