For the first time, more businesses are paying for Anthropic's Claude than for OpenAI's products - at least according to one of the better-sourced datasets in the industry.
Ramp's May 2026 AI Index shows Anthropic ahead in verified business customers. Ramp is a corporate expense and financial management platform used by thousands of companies. Unlike surveys that ask employees which AI tools they prefer, Ramp's data reflects actual commercial transactions - companies paying invoices to specific AI providers. That makes it a more reliable signal than self-reported usage polls.
What the Ramp Index Actually Captures
The index doesn't cover every business - only companies using Ramp for expense management - but it's a consistent, transaction-based sample that's tracked AI spending since the category took off. When a company pays an Anthropic bill through Ramp, it registers as a verified customer. This is the first month Anthropic has registered more of those than OpenAI.
Why Enterprise Buyers Are Choosing Claude
A few things have driven Claude's enterprise adoption. Anthropic's public positioning around AI safety resonates with legal, finance, and healthcare buyers who need to justify AI adoption to boards and regulators. Claude's 200,000-token context window - long enough to process a 500-page document in a single request - has made it practical for document-heavy workflows that smaller context windows couldn't handle reliably.
Coding performance has also brought in enterprise developers, particularly since the Claude Code launch earlier this year. Software teams that started using it for individual developer productivity have pushed their companies toward Anthropic API contracts.
OpenAI isn't disappearing from enterprise sales - ChatGPT still dominates consumer mindshare, and many corporate API integrations predate this shift. But when businesses are actively choosing an AI provider in 2026, Anthropic is winning more of those decisions.
One caveat: Ramp's dataset only includes companies using Ramp specifically. Enterprises managing their own payment systems, or using resellers, wouldn't appear. But the directional shift is meaningful - and the timing coincides with OpenAI's ongoing pricing adjustments and model-tier complexity, which some enterprise procurement teams have found difficult to manage. Anthropic's simpler tier structure (Haiku/Sonnet/Opus) and consistent API behavior appear to be working in its favor.