Anthropic and OpenAI Race to IPO — Founders Built on Them Inherit the Price War

The loudest part of two companies racing to IPO isn’t the stock market. It’s the price sheet. As Anthropic and OpenAI climb onto the same track, the fight between them has started pulling down — or locking up — the costs of every founder building a product on top of them.

What Happened

Anthropic pulled in $65 billion in its Series H, landing at a $965 billion valuation right before going public. Altimeter, Greenoaks, Dragoneer, and Sequoia led the round, and memory suppliers Samsung, SK Hynix, and Micron put their names on the list too. That figure tops OpenAI, which raised $122 billion at an $852 billion valuation just months earlier. Both labs have filed to go public — Anthropic a draft S-1 in early June, OpenAI a confidential S-1 shortly after — and both are pointed at a near-term listing.

What grew the numbers is the enterprise. Anthropic’s annualized revenue jumped from $9 billion at the end of 2025 to $47 billion in May, more than five times in five months, pulled by its coding agent Claude Code. On Ramp’s AI spending index, more companies now pay Anthropic than OpenAI for the first time, and ChatGPT’s share of generative-AI web traffic slid from 77.6% to 53.7% in a year. With share slipping, reports surfaced that OpenAI is weighing cuts to token pricing. At the same time, both companies started attaching mechanisms to keep their year-long enterprise customers locked at current prices. Prices fall, but the right to those prices gets bolted down in a contract. This is two giants on the eve of an IPO colliding head-on over market share.

What This Means for Founders

The first-order effect is good news. When two suppliers cut prices to win share, the inference cost for the founder building on top of those models falls with them. Cheaper tokens mean you can switch back on the features you shelved because the unit economics didn’t work. The second-order effect is the problem. The harder the price war gets, the more both companies move to bind customers — year-long deals, locked current rates, and switching programs that subsidize the cost of migrating. OpenAI’s offer of two free months of Codex plus a one-click tool that ports Claude prompts, skills, and MCP configs for enterprises leaving Anthropic is exactly that signal. Lock-in and poaching happen at the same time.

What a founder faces here is platform risk. If nearly all of your revenue rides on a model bound to one of two companies, that company’s pricing policy, terms, and priorities decide your margin and your roadmap. After the IPO, that pressure sharpens. A public company watching its quarterly numbers can’t subsidize money-losing tokens forever — which means today’s aggressive cuts aren’t permanent. In a market where two companies collide head-on carrying $47 billion in revenue and an $852 billion valuation, model supply is effectively concentrated in two hands. Sink your code, prompts, and operational know-how deep into one of them, and all the leverage in the next price negotiation sits on the other side of the table.

What You Can Do Now

Abstract your model-call layer. If your prompts and tool definitions are wired directly into one provider’s SDK and format, you can’t switch when rates rise or terms change. A routing layer that lets you swap models on cost and quality becomes a negotiating card in itself. When you sign an enterprise contract, don’t reflexively love a locked current rate — in a falling-price phase, that lock can pin you to an expensive number. When cut reports are circulating, a short-term deal that lets you re-sign at a lower rate may beat a long one. And read both suppliers’ IPO calendars and earnings guidance as a cost signal. The rates that came from a private company willing to eat losses can climb again the moment it becomes a public one chasing quarterly results.