Leaked Anthropic IPO Filing Highlights $4.6B Revenue and Deep Cloud Reliance
A confidential IPO filing revealed that Anthropic generated nearly $4.6 billion in revenue in 2025—a 12-fold increase year-over-year—while its operating losses more than doubled to over $8 billion. The documents show that 47% of its sales were routed through cloud partners Amazon and Google, reflecting a growing structural dependency on tech incumbents. The disclosures offer a rare financial benchmark for frontier AI labs seeking trillion-dollar valuations while burning billions in capital. They highlight how major AI developers rely on cloud titans for distribution networks and computing power, creating intricate circular relationships of investment, supply, and direct competition. Usage-based billing accounted for $3.8 billion of Anthropic's revenue, and its long-term compute commitments exceeded $417 billion for 3.5 GW of dedicated capacity by early 2026. Anthropic accounts for cloud marketplace sales on a gross basis (paying ~16% distribution fees), an accounting choice that competitor OpenAI has reportedly claimed inflates top-line revenue.
## BACKGROUND
Anthropic is an artificial intelligence safety and research company founded by former OpenAI executives, best known for developing the Claude family of large language models. Training and running state-of-the-art AI models requires immense compute power, prompting startups to form multi-billion-dollar alliances with cloud infrastructure providers such as Amazon Web Services and Google Cloud.