U.S. enterprises are shifting their focus in AI tool procurement. Ramp’s May AI Index shows that Anthropic’s Claude has risen to a 34.4% adoption rate among U.S. businesses, slightly surpassing OpenAI’s ChatGPT at 32.3%. This data, derived from corporate card and invoice records of over 50,000 U.S. companies, reflects actual spending rather than survey responses.
Growth rates have significantly diverged within a year.
Over the past year, enterprise adoption of Claude increased by approximately fourfold, while ChatGPT saw only a modest 0.3% growth 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 exceed this level, indicating that many companies are not choosing just one vendor but are instead purchasing services from both types of models simultaneously.
Multi-model deployment has become the norm.
According to Ramp’s estimates, approximately 16% of U.S. businesses pay both Anthropic and OpenAI. In other words, about one-third of AI-using enterprises have adopted a multi-model architecture.
This deployment approach better aligns with actual enterprise software usage patterns. Enterprises typically assign models based on specific tasks—for example, using one model for document processing, code generation, or backend workflows, and another for creative content or customer-facing scenarios.
New projects favor Claude.
The article notes that enterprise teams tend to adopt Claude as their default starting point when launching new projects, particularly in software development and coding scenarios. Even though some companies continue to use OpenAI products in other areas, new projects are increasingly prioritizing integration with Anthropic.
This change is related to enterprise needs. Enterprises prioritize model stability in production environments, long-context handling, and consistent instruction following over demonstration effects. These capabilities determine whether an AI system can operate continuously with minimal human intervention.
The focus of procurement has shifted toward implementable capabilities.
The report suggests that corporate adoption of AI is no longer limited to pilot stages but is now more focused on usability and maintenance costs after system deployment. As AI gradually integrates into operational processes, stability and predictability are becoming more important than isolated demonstration results.
However, Ara Kharazian, Chief Economist at Ramp, also cautioned that the market is still in its early stages, and the leading position may continue to shift. Computational constraints, reliability issues, and cost pressures from token-based billing remain factors that enterprise procurement teams must evaluate.
He advises businesses to remain flexible in model selection, prioritize testing platform performance against real production workflows, and avoid prematurely locking infrastructure and contracts into a single vendor.