What Anthropic's IPO tells anyone building with AI

Anthropic, the company behind Claude, is planning to go public on Nasdaq before Thanksgiving. Investors are expected to value it at over $2 trillion — among the largest valuations any company has ever sought. That number has a real story behind it, and the story is not just about Anthropic.

In 2025, Anthropic made $4.6 billion in revenue — twelve times what it made the year before. Claude is now used in more than a third of all U.S. business AI spending. The growth numbers are extraordinary. But costs still outpaced the revenue — the company lost over $8 billion last year.

What makes this unlike a typical tech listing: Anthropic has already locked in $518 billion in future computing costs, and 80% of that cannot be cancelled. Keeping top AI models running means reserving huge amounts of hardware — chips, servers, electricity — years in advance. Closer to financing a power grid than a startup. The public markets are now asked to fund that gap.

The filing is also unusually honest about risk. The U.S. government stopped federal agencies from using Anthropic's models earlier this year. Two newer models were banned from export. The filing itself says advanced AI could pose "catastrophic or existential risks to humanity." These are not standard legal disclaimers — they are signals that AI is entering regulated-industry territory, like energy or finance.

If you build on Claude or any major AI API, this IPO changes something real. A publicly listed Anthropic will answer to shareholders every quarter. That creates pressure on pricing, access tiers, and enterprise priorities. The practical move now: know which AI services your product depends on, check that you can switch if needed, and watch how pricing evolves in the year ahead.