Anthropic Passes OpenAI in B2B: What the Ramp Data Really Says
Agentic AIAnthropic has passed OpenAI in paid B2B adoption for the first time, reaching 34.4% of companies tracked by Ramp’s AI Index versus OpenAI’s 32.3% — a narrow lead that understates a more significant divergence in how the two providers are being used inside enterprise AI stacks.
What the Ramp Data Actually Measures
The Ramp AI Index tracks what share of companies in Ramp’s dataset are paying which AI providers — not volume of usage, API calls, or total spend. Ramp’s customer base skews heavily toward US companies, meaning the percentages do not reflect global enterprise adoption patterns. Both of these caveats matter for interpreting the headline number.
What the data captures well is adoption momentum. Anthropic’s B2B penetration quadrupled over the past year while OpenAI’s grew by just 0.3%. A one-month lead is not a moat, but a 4x growth trajectory over 12 months against a competitor with near-zero growth in the same period is a structural signal — not a statistical noise event. According to Ramp economist Ara Kharazian, the shift reflects agentic model performance driving higher adoption in verticals where Claude’s coding and reasoning capabilities have become the default.
Where the Gap Is Largest
The aggregate adoption percentages obscure the segment distribution. In code generation, Anthropic’s Claude has captured 42% to 54% of global market share according to external analyses, while OpenAI holds approximately 21% of that segment. In enterprise agent deployments, Anthropic leads at 40% versus OpenAI’s 27%.
Code generation is the segment where enterprise buyers most readily switch providers: it is evaluable, the feedback loop is fast, and the quality difference between models is directly observable in CI/CD metrics. Claude’s performance on code generation and agentic task completion — the workflows where Claude Code and multi-agent orchestration are deployed — appears to be the primary driver of the adoption gap in B2B.
For the enterprise AI platform buyer, this matters because code generation adoption is a leading indicator of broader Claude platform adoption. Teams that default to Claude for code reviews, pull request drafts, and test generation tend to expand Claude usage into adjacent workflows — documentation, incident analysis, technical specification drafting — faster than teams using a mixed-model approach.
The So What
The headline obscures both the size of the lead (thin) and the durability concerns. Ramp’s data identifies three headwinds for Anthropic: pricing pressure from cheaper open-source inference alternatives, reported user complaints about outages and quality consistency, and the Opus 4.7 pricing structure that triples image processing costs — a constraint for multimodal agentic applications.
For platform architects evaluating model provider strategy, the Ramp data’s practical implication is not “switch to Claude” but “evaluate the code generation segment independently.” The 54% code gen share suggests Claude’s performance advantage in that workflow is measurable enough that enterprise teams are choosing it over OpenAI on merit, not vendor relationship. The question for every engineering platform team is whether their own model evaluation data reflects this pattern — and whether their current provider mix is optimized for the workflows where the performance gap is largest.
Photo by Tara Winstead on Pexels
Source: The Decoder
Content created with AI assistance and reviewed for accuracy.
Join the conversation
Stack Insiders is our free community for readers who want to go deeper — share resources, ask questions, and connect with others across every vertical we cover.
Join Stack Insiders →