U.S. companies are shifting their purchasing priorities when selecting AI tools. Ramp's May AI Index shows that Anthropic's Claude has seen adoption among U.S. companies rise to 34.4%, slightly higher than OpenAI ChatGPT's 32.3%. This data comes from corporate cards and invoice records of over 50,000 U.S. companies, reflecting actual spending rather than surveys.
Significant divergence in growth rates within a year
Over the past year, Claude's enterprise adoption rate has increased approximately fourfold, while ChatGPT has only seen a slight increase of 0.3% during the same period. This means that in the enterprise AI market, Anthropic has transitioned from a follower to one of the leaders.
Ramp's data also shows that the current overall enterprise AI adoption rate is 50.6%. The combined adoption rates of Claude and ChatGPT are higher than this level, indicating that many companies are not choosing just one vendor, but are purchasing both types of model services simultaneously.
Multi-model deployment has become the norm.
According to Ramp's estimates, approximately 16% of US companies pay for both Anthropic and OpenAI. In other words, about one-third of AI-using companies have already adopted a multi-model architecture.
This type of deployment is closer to the actual usage habits of enterprise software. Enterprises will allocate models according to tasks, for example, using one model for document processing, code generation or backend processes, and another model for creative content or customer-facing scenarios.
The new project is more inclined towards Claude.
The article mentions that enterprise teams tend to use Claude as the default starting point when launching new projects, especially in software development and coding scenarios. Even though some companies still use OpenAI products in other business areas, new projects are starting to prioritize Anthropic.
This change is related to enterprise needs. Compared to demonstration effects, enterprises value the model's stability in a production environment, its ability to handle long contexts, and its consistency in executing instructions. These capabilities determine whether an AI system can operate continuously with minimal human intervention.
Procurement focus shifts to implementation capabilities
The report suggests that enterprises are no longer confined to the experimental stage when purchasing AI, but are now focusing more on the availability and maintenance costs after the system goes live. As AI gradually integrates into operational processes, stability and predictability are becoming more important than the effectiveness of a single demonstration.
However, Ramp's chief economist, Ara Kharazian, also cautioned that the market is still in its early stages, and the leading positions may continue to change. Computing power constraints, reliability issues, and the cost pressures of token-based billing remain factors that enterprise procurement teams need to evaluate.
He advised companies to be flexible in their model selection, prioritize testing platform performance based on real production processes, and avoid prematurely binding infrastructure and contracts to a single supplier.
