Anthropic has overtaken OpenAI in paid U.S. business adoption for the first time, according to the May 2026 release of the Ramp AI Index. 34.4% of businesses on the fintech's platform now pay for Anthropic products. OpenAI sits at 32.3%, having slipped 2.9 percentage points in April.
A year ago this wasn't close
In May 2025, around 9% of Ramp's business users were paying Anthropic. The company has roughly quadrupled that figure in twelve months. OpenAI, over the same window, grew its share by 0.3 percentage points. Not a typo.
These numbers come with footnotes the size of the chart. Ramp tracks corporate card and bill-pay data across more than 50,000 U.S. businesses, which means the index counts paid subscriptions, not token volume, not API calls, not how often anyone actually uses what they're paying for. Free ChatGPT users, who likely outnumber paid Claude users by some margin, don't show up at all.
The methodology is half the story
Ramp's customer base skews venture-backed, tech-forward, and concentrated in information, finance, and professional services. The kind of buyer who was probably going to pick Claude eventually anyway. Ramp economist Ara Kharazian acknowledges this directly: he says the results "should not be construed to suggest Anthropic is the definitive leader in business adoption," which is the sort of thing you write when the headline above your own analysis says exactly that.
What the data can support is the trajectory. Anthropic's growth has been steady, month over month, across multiple Ramp releases. That's harder to dismiss than any single month's snapshot.
Three problems Anthropic can't outrun
Kharazian flags three headwinds in the same post. The first is structural. Anthropic charges by the token, which incentivizes pushing customers toward expensive models even when cheaper ones would do. Uber's CTO recently told The Information the company already blew through its 2026 AI budget.
Reliability is the second. Users have complained about Claude getting worse in recent weeks, with frequent outages and rate caps. Anthropic reset usage limits in April and signed a SpaceX compute deal to relieve the immediate pressure. Whether that fixes the underlying problem before customers start routing around it is a separate question.
The third is a stranger choice. Anthropic's latest model update reportedly triples token costs for any prompt containing an image. The cost-and-compute crunch is the company's biggest known weakness, and the roadmap responds by making one of the most common prompt types more expensive.
OpenAI's Codex does much of what Claude Code does, costs less per task, and runs on infrastructure that isn't currently being patched in public. Switching costs between providers are close to zero.
The next Ramp AI Index lands in June. Kharazian says he'll be watching Codex subscription growth and the rise of AI inference platforms serving cheaper open-source models.
