OpenAI Gains on Anthropic in Enterprise Spending Race
New corporate spending data from Ramp shows OpenAI accelerating in Q3, narrowing the enterprise market share lead Anthropic built earlier this year.
OpenAI is closing the gap with Anthropic in enterprise AI spending, according to new data from expense management platform Ramp reported by TechCrunch. After ceding ground to its rival during the second quarter, the ChatGPT maker is now accelerating faster than Anthropic among corporate buyers in the third quarter.
Ramp tracks spending across more than 70,000 U.S. businesses that manage billing through its corporate cards and spend management software. Its customer base tilts heavily toward technology companies and startups. The figures reflect market share percentages rather than absolute dollar amounts, which Ramp declined to disclose.
Anthropic first overtook OpenAI on the Ramp platform in May, capturing 41 percent market share against OpenAI's 39 percent. By July, Anthropic had stretched that lead to nearly 44 percent while OpenAI held roughly 40 percent. Third-quarter data now shows OpenAI reversing that trend, growing faster among these corporate users than its rival.
Ramp economist Ara Kharazian attributed OpenAI's resurgence to strong developer adoption of its newer GPT-5.6 Sol model. Anthropic, meanwhile, ran into resistance over its higher-tier Fable 5 model: Kharazian cited steep pricing and a regulatory data retention requirement that obligated businesses to keep customer data for 30 days. That condition triggered pushback from enterprise developers with confidentiality concerns.
Overall corporate appetite for AI tools continues to grow regardless of who leads the share count. Ramp reported that the share of its customer base paying for AI tools reached nearly 56 percent in July, up from 50 percent in March. Both companies are expanding revenue as the total pool of enterprise AI spending widens.
The Cognarah Angle
The rapid swings in market share between OpenAI and Anthropic reveal a serious structural problem in enterprise AI: model loyalty is exceptionally thin. For all the talk of proprietary data pipelines and tailored workflows, corporate buyers are treating foundation models like interchangeable cloud compute. When a lab ships a marginally better or cheaper model, technical teams reroute their API calls almost overnight.
That zero-moat dynamic should concern the venture investors pricing these labs at enormous valuations. Traditional software companies command premium multiples because their products embed deeply into enterprise processes, creating friction that keeps customers in place. Generative AI APIs do not share that natural lock-in. A configuration change can redirect developer queries from Claude to GPT in minutes, with no migration project, no retraining, and no contract negotiation.
Anthropic's stumble over mandatory data retention makes the picture sharper. Enterprise buyers will not trade compliance certainty for benchmark performance, regardless of how impressive a model scores on internal evals. That is not a temporary preference; it is a purchasing reality that both companies need to engineer around, not around it. As OpenAI and Anthropic each position for potential public offerings, quarterly share shifts driven by developer sentiment will test investor patience far more than founders would like to admit.
If multi-billion-dollar enterprise AI contracts can evaporate with a single model update, investors should ask whether they are funding durable software franchises or simply subsidizing an expensive compute war with no clear winner.
Reporting sourced from TechCrunch. Analysis and Cognarah Angle are Cognarah's own.
Written by
StacyAI-assisted news curation. Every story is reviewed by our editors before publication.
