OpenAI Weighs Steep Token Price Cuts to Fend Off Anthropic as the Pre-IPO Token War Escalates

OpenAI is mulling sharp reductions to what it charges for its AI offerings as it braces for a price fight with Anthropic, according to a report from The Wall Street Journal. The company is weighing significant cuts to its token pricing — the per-unit metering that AI firms use to bill for model usage — in anticipation of similar cuts it expects Anthropic to make. The deliberations are described as early, with nothing decided.

The logic is a classic volume play: drop the per-unit price, win more usage, and bet that scale eventually makes up for thinner margins. OpenAI appears to view its pricing as a competitive soft spot against Anthropic's enterprise momentum, and the company is reportedly willing to trade near-term revenue for market share. That is a costly bet to make from where OpenAI sits — by some estimates its cash burn runs many times that of its rival, and it does not project profitability until around the end of the decade, roughly two years later than Anthropic's expected 2028 breakeven.

The backdrop is a valuation race that has flipped the industry's pecking order. Anthropic closed its Series H on May 28 at a roughly $965 billion valuation, edging past OpenAI, which was last marked at $852 billion in March. Both companies have now confidentially filed for IPOs within days of each other, turning pricing into something more than an operating decision — whoever locks in the most enterprise share before going public stands to command the richer multiple when the books open.

Pricing pressure is already visible elsewhere on the board. Last week Google cut its AI Plus consumer plan to $4.99 in the US, and ChatGPT became the first app to cross a billion monthly users in May. With consumer subscriptions, API tokens, and frontier-model performance all converging, the marginal cost of intelligence is being pushed down on multiple fronts at once.

For developers building on these platforms, a token-price war is the rare moment when the incentives of the labs and their customers point the same direction. The open question is whether the cuts are deep enough to change build-versus-buy math — or merely a defensive reflex dressed up as generosity. Either way, the era of paying a premium simply because a model carried the OpenAI or Anthropic name looks like it is ending.