OpenAI gains on Anthropic in business market share, Ramp data shows

Until OpenAI and Anthropic get close enough to their planned IPOs to release financials, outside observers have to rely on secondhand signals to gauge how their businesses compare. Ramp, the corporate credit card and expense management company, has published one such signal: new data showing OpenAI clawing back ground against Anthropic among U.S. business customers.

OpenAI was once the runaway leader with both businesses and consumers, but it lost the lead among Ramp's paying business customers back in May. That month Anthropic hit 41% market share to OpenAI's 39%, and OpenAI has not regained the lead since. By July the gap had widened further: Anthropic held nearly 44% to OpenAI's nearly 40%.

The numbers come from a dataset covering more than 70,000 American businesses that together spend billions of dollars through Ramp's bill-pay and corporate-card products. Ramp's customer base spans many industries but skews toward tech, since it is a popular Silicon Valley corporate card; the company also declined to share actual dollar figures, releasing only percentages. TechCrunch notes the numbers are not a measure of the total market either, since they exclude large enterprises that manage spend through other providers such as American Express.

Despite Anthropic's larger and growing lead through July, Ramp economist Ara Kharazian says a closer look at the most recent data shows OpenAI is currently growing faster than Anthropic among this business segment so far in the third quarter. A month of the quarter still remains, and the trend could shift again before it closes. On X, Kharazian credited OpenAI's gains to its newer model, writing, "GPT-5.6 Sol is really good, increasingly the choice for developers," and added that "Fable 5, meanwhile, disappointed both in adoption and real-world application given price + data retention requirements imposed by regulators."

TechCrunch pushes back on that last point, calling it a possible oversimplification. Fable is Anthropic's higher-end model tier, and while it is expensive, it is also built for a more targeted set of use cases than a general-purpose chatbot rather than competing head-on with it. The article also notes that Anthropic caused some outrage when it warned Fable users that it must retain their data for 30 days.

Ramp's data suggests the back-and-forth is happening against a backdrop of overall growth: both companies should be gaining business revenue even as they fight over market share, because the pool of customers paying for AI at all keeps expanding. The share of Ramp's business customers paying for some AI product topped 50% in March and reached nearly 56% by July. The article frames the volatility between OpenAI and Anthropic as a sign that Anthropic hasn't won permanently, and that both companies' investors should be cautious about assuming enterprise AI spending is "sticky."

Key facts

Anthropic overtook OpenAI among Ramp's paying U.S. business customers in May, 41% to 39%, and had widened that lead to nearly 44% versus nearly 40% by July.
Ramp's dataset covers more than 70,000 U.S. businesses spending billions through its cards and bill-pay tools, skewed toward tech companies and excluding large enterprises that use other providers such as American Express.
Ramp economist Ara Kharazian says OpenAI is growing faster than Anthropic among this business segment so far in the third quarter, though a month of the quarter remains.
Kharazian linked OpenAI's gains to its GPT-5.6 Sol model and said Anthropic's Fable 5 tier disappointed on adoption, citing its price and a data-retention requirement; TechCrunch calls that explanation a possible oversimplification.
The share of Ramp's business customers paying for any AI product grew from more than 50% in March to nearly 56% by July, even as OpenAI and Anthropic keep swapping ground.

Why it matters

Neither OpenAI nor Anthropic has gone public, so outside observers still have no official financials to compare the two companies by. Ramp's spending data is one of the few third-party windows into that comparison, and it shows the contest for U.S. business customers is far from settled. Anthropic overtook OpenAI in May and had extended its lead by July, but Ramp's own economist says OpenAI's growth has picked up again in the current quarter. That back-and-forth matters beyond bragging rights: the article reads it as evidence that enterprise customers are still willing to switch labs as new models ship, which should give both companies' investors pause about how locked-in either company's enterprise revenue really is.

Who it affects

U.S. businesses that pay for OpenAI's or Anthropic's products through Ramp's cards and bill-pay tools are the group the data actually measures, more than 70,000 companies skewed toward tech. OpenAI and Anthropic themselves are affected as they compete for that segment ahead of their planned IPOs, and so are investors trying to size up either company's enterprise traction before financials become public. Anthropic's Fable users are affected directly too, since Kharazian ties the tier's softer adoption to its cost and to a data retention requirement imposed by regulators.

How to use it

Treat these percentages as one directional signal, not a market total. They come from Ramp's own base of roughly 70,000 U.S. businesses, a group skewed toward tech companies, and they exclude large enterprises that route spend through other providers such as American Express. Ramp also released only shares of the market, not dollar amounts, so the figures alone show how the two companies' piece of Ramp's pie is shifting, not how much revenue either is actually booking.

How solid is it

The market-share numbers come from Ramp's own transaction data across a named base of more than 70,000 businesses, and the growth comparison is attributed to a named Ramp economist, Ara Kharazian, who posted supporting comments on X. But Ramp declined to release the dollar amounts behind the percentages, and the article itself flags that the dataset does not cover the total market. TechCrunch also pushes back on Kharazian's own explanation for Fable's softer adoption, calling it a possible oversimplification, which suggests some caution is warranted even around Ramp's own economist's reading of the numbers.

Risks and caveats

The data excludes large enterprises that manage spend through providers other than Ramp, so it says nothing about the biggest corporate AI budgets. Only percentages were released, never dollar figures, so there is no way to gauge the actual size of either company's business revenue from this alone. The finding that OpenAI is growing faster in Q3 covers a quarter that has not finished yet, and the article explicitly warns the trend could reverse again before it ends; no figures beyond "growing faster" have been published to back the claim. Kharazian's attribution of Fable's weaker adoption to price and a 30-day data retention requirement is also disputed by the article itself as an oversimplification, since Fable is described as a targeted, higher-end tier rather than a general-purpose chatbot competitor.

“GPT-5.6 Sol is really good, increasingly the choice for developers”

— Ara Kharazian, Ramp economist, on X