Enterprise AI spending may be growing fast, but loyalty is another story. According to TechCrunch, fresh data from corporate expense platform Ramp shows that OpenAI has been losing ground to Anthropic among U.S. business users since May, but is now growing faster in Q3. That flip-flopping pattern is arguably more interesting than who is currently ahead.
Ramp tracks spending across more than 70,000 American businesses that use its corporate card and bill pay products. In May, Anthropic overtook OpenAI in market share among those businesses, hitting 41% to OpenAI’s 39%. As of July, Anthropic has extended that lead to nearly 44% versus OpenAI’s nearly 40%. But Ramp economist Ara Kharazian noted that OpenAI is currently growing faster in Q3 so far, suggesting the gap could narrow again before the quarter ends.
There are real caveats here. Ramp’s customer base skews toward tech-oriented companies and startups, particularly those in Silicon Valley. Large enterprises using spend management tools from providers like American Express are not captured in this data. So this is a directional signal, not a census. Still, 70,000 businesses spending billions is enough to take seriously, especially when more complete financial data from either company remains unavailable ahead of their respective IPOs.
Kharazian pointed to OpenAI’s GPT-5.6 Sol as a driver of its recent momentum, describing it as increasingly the choice for developers. On the Anthropic side, he flagged that Fable 5, the company’s higher-end model tier, underperformed on adoption. Part of that is pricing, and part of it is a data retention policy that requires Fable users to accept 30-day data retention, a requirement that generated real pushback when it was announced. Anthropic would argue Fable is built for specific, higher-stakes use cases rather than general-purpose chat, which makes broad adoption comparisons a bit unfair. But perception matters, and the backlash was real.
What this data actually reveals is that enterprise AI spending has very little stickiness right now. Businesses are switching between providers as new models drop, which means neither OpenAI nor Anthropic has built the kind of deep integration lock-in that typically defines enterprise software. That should concern investors in both companies. If customers move freely based on whichever model performed best last month, then maintaining market share requires a relentless release cadence rather than durable product advantages.
The broader trend is more encouraging for both. The share of Ramp customers paying for AI at all has been climbing steadily. It crossed 50% in March and reached nearly 56% by July. So the market is expanding even as the two leading players trade positions at the top. That means both companies can show revenue growth even during periods when they are losing relative share.
For developers and technical buyers evaluating which platform to commit to, the instability in these rankings is a reasonable argument for staying flexible. Building deep dependencies on either provider’s API carries real switching costs if pricing, data policies, or model quality shifts. Right now the data suggests those shifts happen often enough to matter.
