Preface

Context: With both OpenAI and Anthropic still some distance from potential initial public offerings, their detailed financials remain private. To gauge how each company is performing commercially, we must rely on alternative signals. One such indicator is recent data published by Ramp, a corporate credit card and expense management provider. This dataset offers a snapshot of AI adoption among U.S. businesses that use Ramp’s payment and expense tools. The purpose of this article is to summarize Ramp’s findings and explain what they suggest about market share dynamics between OpenAI and Anthropic, as well as broader trends in enterprise AI spending.

Lazy bag

Key takeaway: Ramp’s data shows OpenAI closing the gap on Anthropic among U.S. businesses that use its platform. Anthropic led earlier this year, but recent growth rates suggest OpenAI is catching up in Q3. The dataset covers tens of thousands of companies and points to a growing share of businesses paying for AI.

Main Body

Until both OpenAI and Anthropic move closer to IPOs and publicly release financial statements, observers must look to alternative sources to assess commercial traction. One valuable proxy comes from Ramp, a corporate expense and bill-pay provider whose customers include more than 70,000 American businesses. Ramp’s latest dataset offers insight into which AI providers these companies are paying for through Ramp’s corporate card and bill-pay services.

Earlier this year, Anthropic seized the lead among Ramp’s paying business users. In May, Anthropic accounted for roughly 41% of paid usage, while OpenAI held about 39%. That lead held and even widened slightly into July, when Ramp reported Anthropic at nearly 44% and OpenAI at nearly 40%. The figures are percentages rather than dollar amounts, because Ramp declined to share absolute spending values.

Despite Anthropic’s earlier advantage, Ramp’s more recent week-to-date and month-to-date analysis suggests a shifting dynamic. According to Ramp economist Ara Kharazian, OpenAI’s usage among Ramp customers has been growing faster in Q3 to date than Anthropic’s. That trend indicates OpenAI is beginning to close the gap, though the quarter was not finished at the time of the report — leaving room for the trajectory to change before Q3 concludes.

It’s important to understand what this sample represents and what it omits. Ramp’s customer base skews toward technology and other industries that commonly use Silicon Valley–style corporate cards and expense platforms. Large enterprises that manage vendor payments through other financial providers, such as American Express, are not included. Accordingly, the data should not be interpreted as a complete measure of overall market share across every enterprise segment. Still, with tens of thousands of firms represented and billions of dollars routed through Ramp’s services, the findings provide a meaningful indicator of market trends among a substantial portion of U.S. businesses.

One implication of Ramp’s numbers is how fluid enterprise purchasing can be in AI. Anthropic’s early lead was not definitive or permanent. As each AI lab releases new models or adjusts pricing and data policies, businesses appear willing to switch providers. That fluidity raises questions for investors about how "sticky" enterprise AI spending will be once initial integrations and trials are complete.

Product differences and policy choices matter. Kharazian pointed to OpenAI’s newer model, described as "GPT-5.6 Sol," as a factor in OpenAI’s recent momentum among Ramp customers. In contrast, Anthropic’s higher-tier model, Fable, saw lower-than-expected adoption in this dataset, according to Ramp’s commentary. Part of the issue may be cost and the specific use cases targeted by Fable; it is a premium option designed for certain enterprise needs rather than broad chatbot use. Anthropic also prompted concern among some customers when it said Fable would retain user data for a set period, a policy that can complicate regulatory and procurement decisions for companies with strict data governance requirements.

Another notable point from the Ramp data is that the overall share of companies paying for AI has been increasing among its customers. In March, just over half of Ramp’s customer base paid for AI services; by July that figure had climbed to nearly 56%. That growth suggests rising enterprise adoption even as the two vendors contest market share.

Because Ramp reported percentages rather than absolute spending, it’s not possible to determine whether one provider is significantly out-earning the other in dollars across the full market. However, rising paid-adoption rates combined with shifting vendor percentages imply that both companies are likely seeing growth in business revenue—though gains may be uneven and subject to short-term swings tied to model releases, pricing changes, and data policies.

For stakeholders, the takeaway is twofold. First, Anthropic currently held a lead in Ramp’s snapshot through mid-summer, but OpenAI showed signs of accelerating growth in Q3. Second, enterprise AI adoption appears to be expanding among Ramp customers overall, even if customers demonstrate volatility in vendor choice. Observers should treat these findings as directional signals rather than definitive measurements of total market share.

As both labs continue to launch models and refine enterprise offerings, monitoring multiple data sources — vendor reports, customer surveys, and payment-platform analytics like Ramp’s — will remain essential for understanding the pace and durability of enterprise AI spending.

Key Insights Table