Sarah Friar described the reversal with a pair of numbers and no fanfare. At the start of the year, the OpenAI chief financial officer said, 60 percent of the company’s revenue came from individual ChatGPT subscribers and the rest from business customers. That structure has now flipped.
Friar, speaking in early September, said enterprise revenue grew 32 percent on an annualized basis in June and July, well ahead of the company’s overall annualized revenue growth of 20 percent over the same period. The shift means OpenAI is increasingly a company that sells to other companies rather than to the public.
The pivot is deliberate. OpenAI is racing Anthropic, its closest rival, for large enterprise accounts. Anthropic is expected to go public as soon as this fall, with a raise that could exceed $100 billion, according to people familiar with the plans. That prospect has pushed OpenAI to sharpen its pitch to chief information officers and procurement teams that buy AI the way they buy software licenses.
Friar’s most consequential change is how those customers will pay. She said OpenAI plans to move away from charging businesses by the token, the basic unit of text a model reads and writes, and toward pricing based on value delivered. The details were left open, but the direction is clear: OpenAI wants to decouple its revenue from raw usage and attach it to outcomes.
The shift matters because token pricing has become a race to the bottom. Rivals have cut prices aggressively, and customers have learned to arbitrage between models. Charging for results instead of volume is an attempt to escape a market where the product is increasingly interchangeable and the only lever is price.
Behind the sales push is a shortage OpenAI cannot easily fix. The company expects to spend roughly $750 billion on computing by 2030, according to a person familiar with the figures who was not authorized to discuss them publicly. Even at that scale, the person said, OpenAI still considers itself badly short of compute.
That constraint shapes everything else. If OpenAI is going to spend three-quarters of a trillion dollars on infrastructure, it needs revenue with the predictability that only contracts bring. Consumer subscriptions are volatile. Enterprise agreements, once signed, renew. The flip in Friar’s revenue split is less a triumph of salesmanship than a necessity of the balance sheet.
Anthropic’s looming IPO adds urgency. A public listing would give the rival a war chest and a currency for acquisitions, and it would force both companies to disclose financials that have so far been kept private. OpenAI, which has raised money at valuations that have climbed into the hundreds of billions, would face a newly legible competitor.
Friar framed the enterprise shift as a natural evolution of a company that started selling to developers. That early audience, she argued, already behaves like business customers: they build products, watch costs, and demand reliability. The corporate push extends the same logic to the Fortune 500, where budgets are larger and switching costs are higher.
The value-based pricing plan carries risk. Customers that buy on outcomes will want those outcomes defined and measured, and a vendor that fails to deliver may owe refunds or lose renewals. Token pricing, for all its flaws, is at least unambiguous. Value pricing puts OpenAI on the hook for results it does not fully control.
Analysts said the transition will take time. Most enterprise contracts today are still written around usage, and renegotiating them around outcomes requires sales teams to make a case they are only beginning to develop. The 32 percent enterprise growth is real, but it is growth off a smaller base than the consumer business.
Friar offered no date for when value pricing would become the default. She did, however, make the company’s priorities plain: sell to businesses, charge for what the technology accomplishes, and spend whatever it takes on compute. For a company that once described itself as a research lab with a consumer hit, the transformation is nearly complete.
OpenAI has built the enterprise business on top of a consumer hit. ChatGPT’s hundreds of millions of users created the brand recognition that now opens doors in the executive suite, and the company has converted that attention into contracts through ChatGPT Enterprise and its developer tools. The flip Friar described is the result of years of that conversion work.
The token-to-value shift also mirrors how software pricing has evolved before. Early cloud services charged by the hour; modern vendors charge by seat, by outcome, or by transaction. OpenAI is walking the same path, betting that customers will pay more for results they can measure than for compute they cannot easily judge.
What remains open is whether the market will pay for results the way OpenAI hopes. The flip in revenue is the first signal. The pricing model is the second. Both point the same direction: OpenAI is no longer betting its future on individual subscribers alone.
