New spending data from corporate card provider Ramp indicates that OpenAI is regaining momentum with US businesses in the third quarter, even as Anthropic continues to hold an overall lead in market share. The figures, released this week, offer one of the clearest public glimpses yet into how enterprise AI spending is shifting between the two frontier labs.
Ramp’s data, which covers more than 70,000 American businesses using its bill pay and corporate card products, shows Anthropic’s share of AI-related spending peaked at nearly 44% in July, compared with OpenAI’s nearly 40%. That gap has narrowed in Q3 to date, with OpenAI growing faster among this segment, according to Ramp economist Ara Kharazian. The dataset spans billions of dollars in transactions and skews toward tech companies, given Ramp’s popularity in Silicon Valley.
Also read: AI accounting startup Rillet raises $100M, hits unicorn status in 48 hours
How Anthropic took the lead — and why it’s slipping
Anthropic first overtook OpenAI among Ramp’s paying business users in May, when it hit 41% market share to OpenAI’s 39%. The shift came as enterprises increasingly adopted Claude for coding and agentic workflows. But the latest trend line suggests the competitive space remains highly fluid, with customers willing to switch as each lab ships new models.
Kharazian pointed to specific product dynamics driving the recent swing. “GPT-5.6 Sol is really good, increasingly the choice for developers,” he posted on X. “Fable 5, meanwhile, disappointed both in adoption and real-world application given price + data retention requirements imposed by regulators.”
Also read: Michael Polansky's Outer Biosciences keeps human skin alive for a month to train AI on real tissue
Anthropic’s Fable tier, positioned as a higher-end offering for complex enterprise tasks, carries a premium price and triggered backlash earlier this year when the company disclosed that it would retain user data for 30 days to comply with regulatory obligations. While Fable is designed for targeted use cases rather than general chatbot interactions, the combination of cost and data-handling concerns appears to have dampened its uptake among Ramp’s customer base.
What the numbers do and don’t capture
Ramp declined to disclose actual dollar figures, sharing only percentages, and the dataset has notable limitations. It excludes large enterprises that use spend-management platforms from providers like American Express rather than Ramp. As a result, the numbers should be read as directional market indications rather than a complete picture of the enterprise AI market.
Still, the data points to a broader trend: overall adoption of paid AI tools among businesses is expanding. The share of Ramp customers paying for AI services topped 50% in March and reached nearly 56% by July, suggesting that both OpenAI and Anthropic can grow revenue even while competing for share.
The volatility between the two labs carries implications for investors in both companies as they move toward planned IPOs. Enterprise AI spending has often been framed as “sticky,” with high switching costs once teams integrate a model into their workflows. Ramp’s data challenges that assumption, showing meaningful shifts in spending within a single quarter as new models launch and trade-offs around price, performance, and data governance come into focus.
With roughly a month left in Q3, Kharazian cautioned that the trend could shift again before the quarter closes. The pace of model releases in AI means a single launch can reset the competitive balance quickly. For businesses evaluating AI vendors, the data reinforces the value of maintaining flexibility rather than locking into a single provider’s ecosystem.
This article is for informational purposes only and does not constitute financial advice. Market conditions are volatile and forward-looking statements involve uncertainty.
