Inside Anthropic’s trillion-dollar AI gamble

Anthropic’s recent S-1 filing reveals the complex financial architecture of a frontier AI lab, highlighting a transition from massive compute expenditure to high-margin inference profitability. Joey Brookhart, an analyst at SemiAnalysis, explains that Anthropic has achieved 85% gross margins on inference by optimizing token throughput across a diverse chip fleet. Despite this efficiency, the company faces significant pressure to balance immediate revenue generation with the capital-intensive pursuit of AGI. The industry remains highly competitive, with OpenAI and open-source models challenging Anthropic’s market share, while hyperscalers like AWS and Google serve as both essential infrastructure providers and potential rivals. Ultimately, the sustainability of these AI businesses depends on their ability to scale beyond coding-specific use cases into broader enterprise verticals, justifying the massive long-term compute commitments required to maintain a competitive edge at the frontier.