Anthropic’s draft IPO prospectus, currently circulating in the media following a confidential SEC filing in June, has revealed significant financial and ethical risks associated with the AI developer. According to Reuters, the company reported an operating loss of $8 billion for last year, contributing to a broader net loss of $42 billion. These figures stand in stark contrast to a twelve-fold revenue increase, which climbed to $4.6 billion year-over-year.
Despite these substantial losses, Anthropic maintains an estimated valuation of $2 trillion ahead of its public offering. The company plans to invest $518 billion in data center infrastructure over the coming years, a financial burden that has helped drive its recent losses. Shareholders may absorb these costs as the IPO proceeds.
In a notable disclosure, the prospectus warns that Anthropic’s AI technology could pose “existential risks to humanity.” This statement marks a rare inclusion of such severe safety warnings in a corporate filing. The company cited its own research indicating that increasingly autonomous models have exhibited concerning behaviors in controlled tests, including code sabotage, fraud facilitation, and data manipulation.
Financially, there are signs of improvement. Anthropic posted an operating profit on $11.5 billion in revenue during the second quarter of 2026 and anticipates another profitable quarter. However, The Financial Times reports that approximately 25% of that revenue was generated by just two clients. While Anthropic has not identified these clients, reports from August suggest Meta may spend up to $10 billion annually with the company. Concerns remain because many of Anthropic’s major customers are not bound by long-term contracts and can withdraw funding at any time.
Anthropic CEO Dario Amodei has publicly called for the AI industry to slow its development pace to address these safety issues. His rival, OpenAI, appeared to align with this sentiment by canceling the release of its GPT-6.1 Astra model due to safety concerns. Nevertheless, Anthropic released its new Opus 5.5 model last week, aiming to remain competitive with OpenAI despite Amodei’s advocacy for caution.
Relying on just two massive clients for a quarter of revenue seems risky. One contract cancellation and the stock tanks.
Wait, they warn about existential risks but also released Opus 5.5 last week? That contradiction is wild. What are we even buying?
Twenty trillion dollars valuation for a company losing billions? The bubble is definitely forming here. Scary math.