Top AI spenders cut costs 10% as token prices drop
top ai spenders cut per-employee costs by 9.7% in august to $7205, driven by falling token prices and a shift to cheaper models.
ramp's ai index for september 2026 shows that the top 1% of ai spenders in the us reduced their per-employee spending by 9.7% in august to $7205. this decrease is largely driven by a 41% drop in token prices since march 2026, which now stand at $0.68 per million tokens. additionally, there is a noticeable shift away from expensive frontier models towards more affordable alternatives like gpt-5.6 terra and claude's sonnet series. the usage of frontier models like opus, fable, and sol decreased from 53% to 45% of all tokens consumed between early august and early september. open-weight models still have a minimal impact, with only 6.4% of ai-using companies on the ramp platform utilizing them.