Ramp data shows OpenAI closing the gap with Anthropic in business AI

Introduction

Until OpenAI and Anthropic disclose fuller financial information, corporate spending platforms offer one of the few outside views into how their business products are performing. New data from Ramp, cited by TechCrunch, suggests that Anthropic remains ahead of OpenAI among U.S. business customers, while OpenAI is beginning to close the gap.

Key findings

• Ramp’s dataset covers more than 70,000 U.S. businesses using its bill-pay and corporate-card products. Those customers account for billions in spending, but Ramp did not provide the actual dollar amounts spent on AI services.
• Anthropic reached 41% of Ramp’s paying business users in May, compared with 39% for OpenAI. By July, Anthropic was close to 44% and OpenAI close to 40%.
• Ramp economist Ara Kharazian said OpenAI was growing faster among this customer segment in the third quarter so far. The quarter was not yet complete, so the direction could still change.
• Ramp’s customers skew toward technology companies, and the sample excludes large businesses using alternative spend-management providers such as American Express. It should therefore not be treated as a measure of the entire market.
• The share of Ramp customers paying for AI rose from above 50% in March to nearly 56% in July, indicating that the overall market is expanding even as vendors compete for share.

Why the lead remains unstable

The most important signal is not simply whether OpenAI retakes first place. It is that business customers appear willing to switch, test multiple providers, or reconsider their existing contracts as new models arrive. Capability, price, data-retention rules, and the fit between a model and a specific workflow can all affect procurement decisions.

Kharazian linked OpenAI’s recent growth partly to the appeal of GPT-5.6 Sol among developers. He also argued that Anthropic’s Fable 5 faced weaker adoption and real-world use because of its price and data-retention requirements. That explanation is useful but incomplete. Anthropic’s higher-end offering is aimed at more targeted, demanding use cases, so a higher price does not automatically indicate weak competitiveness. The company also drew criticism after warning some Fable users that their data would need to be retained for 30 days.

For enterprises, the issue is less about benchmark leadership than about the trade-off among performance, compliance, privacy, and predictable cost. A model that is excellent in a narrow workflow may still lose a broader purchasing decision if its commercial or governance conditions are difficult to accept.

Implications

The Ramp figures suggest that neither company has permanently secured the enterprise market. A new model release can shift adoption quickly, which should make investors cautious about treating enterprise AI spending as highly “sticky” in the way traditional software subscriptions often are.

At the same time, the rising share of businesses paying for AI is a positive sign for both labs. They may be growing revenue even while competing for relative share. The next question is whether model improvements can become durable workflow gains, stronger data governance, and costs that businesses can forecast over time—not merely a temporary change in quarterly user percentages.

Source: TechCrunch AI