OpenAI Narrows Anthropic’s Business AI Lead, New Ramp Data Shows

Ramp, the corporate credit card and expense management company, has released new data showing OpenAI growing faster than Anthropic among U.S. business customers in the current quarter, even though Anthropic still holds the larger overall share of enterprise AI spending. The numbers, reported by TechCrunch on August 20, are one of the few outside signals into how the two labs are actually performing with paying business customers, since neither OpenAI nor Anthropic has released financials ahead of its planned IPO.

OpenAI was the runaway leader with both businesses and consumers until May 2026, when Anthropic first took the lead among Ramp’s paying business users, 41 percent to OpenAI’s 39 percent. OpenAI never regained that lead. By July, according to Ramp’s own AI Index published August 12, Anthropic held 43.5 percent of business adoption (up 1.1 percentage points from the prior month) to OpenAI’s 39.7 percent (up 0.23 points). But in a more recent look at the numbers, cited by TechCrunch, Ramp economist Ara Kharazian said “OpenAI is currently growing faster among this segment in Q3 to date than Anthropic.” TechCrunch cautioned that a month still remains in the quarter, calling that stretch “like 30 AI years” given how quickly the market has moved, so the trend could shift again before the quarter closes.

The data draws on more than 70,000 American businesses that spend through Ramp’s bill pay and corporate card products. TechCrunch noted the sample skews toward the tech industry, since Ramp is a popular corporate card among Silicon Valley companies, and that it excludes larger enterprises that route spending through other providers such as American Express. Ramp also declined to share actual dollar figures, providing only percentages.

Kharazian pointed to model-level dynamics behind the shift. “GPT-5.6 Sol is really good, increasingly the choice for developers,” he wrote on X, as quoted by TechCrunch. “Fable 5, meanwhile, disappointed both in adoption and real-world application given price + data retention requirements imposed by regulators,” he continued. Ramp’s own AI Index backs up the adoption gap: Fable 5, Anthropic’s higher-end model tier, accounts for only 6 percent of tokens businesses purchase from Anthropic, versus 25 percent of OpenAI’s tokens going to GPT-5.6 Sol. Fable 5 is priced at roughly $10 per million tokens, about twice GPT-5.6 Sol’s roughly $5 per million, and despite that premium it draws only about 75 percent as much model-attributed spend as Sol does.

TechCrunch noted that framing may be an oversimplification: Fable is built for a narrower set of higher-end use cases than a general-purpose chatbot, not a straight swap-in for Sol. Still, Anthropic drew some criticism when it told Fable users it must retain their data for 30 days.

The broader enterprise AI market kept expanding regardless of who is winning share. The portion of Ramp’s business customers paying for AI services topped 50 percent in March and reached nearly 56 percent by July. Use of model serving platforms, services offering open source models and some Chinese-developed models, also edged up to 6.1 percent of AI-paying businesses in July, a gain of 0.2 points from the prior month.

In his own Econ Lab newsletter, Kharazian framed the pattern less as either lab winning outright and more as businesses hitting a spending ceiling. “So with Fable 5, we’ve found a new upper bound for how much businesses are willing to spend on AI,” he wrote. “Here, more performance is not worth the price tag.” For OpenAI and Anthropic investors, the more unsettling takeaway may be the volatility itself. Businesses are willing to switch labs as each releases new models, a sign that enterprise AI spending could be less “sticky” than either company would like heading into an IPO.