OpenAI Wants A Price War With Anthropic—Is It Proving DeepSeek Right?

OpenAI is considering reducing the prices it pays developers and businesses, on the Wall Street Journalexpecting similar cuts from Anthropic. Negotiations are said to be ongoing as both companies have privately filed for IPOs this month, and none have been profitable.

“I think we’re going to have a lot of ways to help people spend less,” said Sam Altman at a recent event. Wall Street Journal. That comment came on the back of OpenAI posting a – 122% adjusted usage limit in Q1 2026 – meaning it lost $1.22 for every dollar it brought in.

Stress is real. Like Decrypt what was said before, Part of ChatGPT the global AI market share fell from 77.6% in May 2025 to 53.7% by April 2026. For the first time, more companies covered by the Ramp AI Index are paying Anthropic than OpenAI. Anthropic’s annual growth is expected to reach $9 billion by the end of 2025 $47 billion by May 2026– a jump of 422% in five months – led by Claude Code, Q2 2026 will be the company’s first profitable quarter.

OpenAI has made its own coding tool, Codex, the company’s most important tool. But it’s playing.

Both companies are fighting a silent battle to attract as many customers as possible amid the world’s biggest tech crisis since the dot-com era. Companies of all kinds are rushing to use AI in one way or another. Uber’s CTO burned through his entire 2026 AI budget by April, some JP Morgan employees have. to waste more in using AI beyond their payroll, according to the bank’s chief data officer for its payroll division.

This is a practice Silicon Valley has adopted to call for “tokenmaxxing”—burning as many AI tokens—pieces of data processed by AI models—as possible, often without a refund. Palantir CEO Alex Karp in comparison to profanity at AIPCon last week. Analysts at JP Morgan published a note this month titled “AI Bills Are Out of Control. “The companies that are facing the biggest problem are the ones that are now thinking about a price war.

Tommy Shaughnessy of Delphi Ventures laid out the trap for the most widely distributed format X post this week: The price of $20/month is always the lowest price that is used the most – a loss guide designed to drive adoption, not counting. When a real business needs AI at scale, it moves to an API, paying for each token, but using a lot of computing power.

Not everyone agrees with this. Some believe that the AI ​​oligopoly in the Western hemisphere allows companies to charge exorbitant prices for processing what they want – Chinese brands pay less for proof. If so, there may be room for more price adjustments while you’re still in the economy.

Real businesses are being sent to API pricing, and companies are burning through cash faster than previously reported fees. Meanwhile, open source providers (companies that provide computing power so that AI models can analyze data) are on the rise, and the tools are a catalyst for their growth. These platforms use leading Chinese AI models such as DeepSeek, GLM, MiMo, Kimi or Minimax, which compete with Claude Opus in benchmarks, in general. about one-thirteenth the value of the closed path.

“Chinese labs open to higher grades,” Shaughnessy he wrote. “This type is the biggest asset an information provider has, and they get it for free.” As long as this continues to work, the bottom of the smart tree will drop to zero – and any recovery from OpenAI or Anthropic will be a mathematical problem with no clear answer.

All of these ideas would end only if China shuts down, Shaughnessy said, which would be a boost to US labs.

Meanwhile, many Chinese AI labs seem to be committed to a different approach.