Anthropic’s business plan: Build God or die trying

Why it matters

Anthropic's IPO filing reveals a company whose survival depends on extraordinary continued growth. The document shows steep revenue expansion and positive adjusted income once accounting charges are set aside, suggesting Anthropic currently earns enough to cover itself. But it also discloses massive long-term, largely non-cancelable commitments for compute and infrastructure, with lease and purchase obligations stretching roughly two decades into the future. The writer frames Anthropic as more disciplined than competitors like OpenAI and Google, having stayed focused on enterprise work such as coding and cybersecurity rather than chasing consumer products, with customers who spend heavily and stick around. Yet Anthropic faces pressure from rival frontier labs, open-source alternatives, and Google's distribution advantages from bundling AI into existing business products. Because revenue tends to spike after each new Claude release, Anthropic can never stop training models. Outside analysts estimate the commitments demand enormous annual spending for years, meaning growth is contractual, not optional. If recursive self-improvement stalls, this looks like a fragile bet.