SIGNAL
Enterprise AI vendor share now turns over by the quarter, not the year. The two leading frontier labs have already swapped the lead once in 2026, and one driver is procurement policy alongside price and model-quality factors. That is the story underneath the monthly horse-race headlines.
Ramp's card-and-bill-pay data from more than 70,000 US businesses shows Anthropic at 43.5% in July, up 1.1 percentage points month over month, and OpenAI at 39.7%, up 0.23 points. Anthropic overtook OpenAI in May at 41% to 39% and has not given the lead back, even as TechCrunch reports Q3-to-date growth tilting toward OpenAI. Ramp's own August letter flags that OpenAI "underperformed overall AI adoption" in the same month.
The mechanism is privacy terms and model-release pacing. OpenAI's August 19 zero-data-retention announcement and Anthropic's reported policy shift on August 20 are competing moves for the same buyer: a corporate legal team that signs a 12-month contract after the privacy clauses change. The causal link between privacy policy moves and share churn is inferred from the temporal clustering of events, not directly observed in the Ramp spend data alone. Kharazian inferred that the volatility in share data challenges the "sticky enterprise AI revenue" thesis underwriters have used to value both labs. The right read of "OpenAI is gaining" is that the lead is up for grabs every quarter, and the next policy memo is worth more than the next benchmark.