Ramp’s AI index finds US companies using more AI and spending less

What the index measures

Ramp AI Index tracks transaction data from more than 70,000 US companies and collects token-spending data from a subset of them. According to Kharazian, its analysis counts API spending only and reflects spending by larger AI customers more heavily.

The decline is not the first, but this one has lasted unusually long. Spending peaked in July.

The gap between use and spend

More AI use alongside lower spending complicates the usual assumption that adoption should show up as a rising bill. The index does not explain the gap, and its API-only view leaves out other ways companies may use AI.

I think the most revealing detail is not the month-over-month decline but how concentrated token spending remains: open models are still below 5%, while Anthropic and OpenAI together account for 95.5% in the final week measured. The index tracks spending, not which models companies use outside paid APIs, so that share is not a complete picture of adoption.

What I’d want to know is whether the longer decline reflects lower prices, more efficient usage, or customers pulling back. The data reported here cannot distinguish among them, leaving the headline’s combination of rising use and falling spend unresolved.