Anthropic's $750 Billion Compute Spending Plan Threatens Profitability

Anthropic's draft IPO filing reveals plans to spend up to $750 billion on AI compute over the next decade, a massive commitment that could significantly delay its path to profitability. The company's heavy investment in computing power for its Claude AI model is a major hurdle for future shareholders. The bulk of the spending includes $160 billion to lease Alphabet processors via Broadcom, $111 billion for Google Cloud, $110 billion for Amazon Web Services, $86 billion for Nvidia semiconductors, and $31 billion for Microsoft Azure. Notably, $450 billion of these cloud contracts are noncancelable, meaning Anthropic must pay regardless of usage, posing a risk if demand slows or competition intensifies. Anthropic's revenue is accelerating, reaching $16.3 billion in the first half of 2026 versus $1.1 billion a year earlier, but losses hit $80 billion, mostly from accounting adjustments. The company was slightly profitable on an adjusted basis in Q3, yet GAAP profitability appears distant, especially with two customers accounting for 20% of revenue.