Skip to main content

Anthropic's IPO Filing: $8B Loss, $54.6B in Compute Commitments

Anthropic has filed its IPO prospectus, and the numbers tell a story of explosive growth paired with deep financial strain. The company, known for its Claude AI models, is aiming for a valuation around $2 trillion. But behind that ambitious target lies a sobering reality: in 2025, revenue surged twelvefold to nearly $4.6 billion, yet operating losses doubled to over $8 billion.

Where the Money Comes From—and Where It Goes

Most of Anthropic's revenue—about $3.8 billion—flows from usage-based billing for the Claude model. Subscription services, by contrast, brought in just $789 million, less than a fifth of total revenue. That reliance on pay-as-you-go access means income is tightly tied to how much customers actually use the model, not to predictable recurring fees.

Image

A significant chunk of that revenue passes through Amazon and Google Cloud. In 2025, Anthropic earned $2.16 billion via these platforms, a share that jumped from 11% in 2023 to 47%. For the privilege, it paid roughly $351 million in channel fees. But here's the twist: Amazon and Google aren't just distribution partners. They're also investors, compute suppliers, and—through their own AI efforts—competitors. That tangled web raises obvious questions about pricing power and potential conflicts of interest.

Customer Concentration and Compute Commitments

Two anonymous customers accounted for 24% of Anthropic's revenue, and neither has a long-term agreement. That kind of concentration makes future income less predictable. Meanwhile, the company has locked itself into $54.6 billion in non-cancellable compute hosting commitments through the end of 2025. Unpaid invoices total $909 million, with 60% collected through intermediaries.

In plain terms, Anthropic has secured the computing power it needs to train and run its models—but at a rigid, massive cost. If demand doesn't keep pace, those commitments could become a heavy anchor.

What This Means for the AI Industry

Anthropic's filing offers a rare glimpse into the economics of a top-tier generative AI company. It shows that even the leaders have scaled commercialization successfully, yet still wrestle with limited control over distribution, inflexible compute costs, and customer churn. The path to profitability remains steep, and the prospectus makes clear that size alone doesn't guarantee stability.

Key Points

  • Anthropic's 2025 revenue hit $4.6 billion, but operating losses exceeded $8 billion.
  • Usage-based Claude access drives most revenue; subscriptions account for under 20%.
  • Amazon and Google are both partners and competitors, handling 47% of revenue.
  • Two unnamed customers make up 24% of revenue, with no long-term contracts.
  • The company faces $54.6 billion in non-cancellable compute commitments and $909 million in unpaid invoices.