Anthropic files for IPO with $104B losses and $33B revenue

Anthropic has filed its IPO prospectus, with the first note scheduled for November 19, marking a critical test of the current AI investment narrative. The document reports that the company accumulated $104 billion in losses since the start of the year, a figure largely driven by non-cash write-downs on convertible debt rather than operational cash burn. Despite this, revenue rose from $2.5 billion to $33.2 billion in the first nine months of 2026, with Q3 revenue hitting $17 billion, a 1,127% year-over-year increase. Operating cash flow improved to positive $6.5 billion, indicating the company is generating liquidity. The prospectus details $895 billion in planned investments for data centers and infrastructure, with $895.4 billion in future compute commitments over five years. Analysts estimate a potential valuation corridor between $1.5 trillion and $2 trillion, though speculative trading on crypto platforms like Hyperliquid has indicated valuations fluctuating between $1.8 trillion and $2.3 trillion. Goldman Sachs is leading the syndicate, with retail access planned through banks including BNP Paribas, Deutsche Bank, and UBS. The filing includes risk disclosures regarding potential existential risks and the acceleration of AI development. Anthropic warns that AI could soon significantly accelerate the development of new models, potentially reaching a threshold where AI development is at least twice as fast with AI support. This could lead to recursive self-improvement, making it harder for humans to control and verify individual steps. The company acknowledges that significant accelerations were often only recognized retrospectively, though no current model has yet reached this specific threshold. The market is shifting from rewarding high AI capital expenditure to questioning profitability, with investors scrutinizing whether AI investments will generate returns.