Ramp AI Index: Top 1% of US Firms Cut AI Spend Per Employee Nearly 10% in August
AI spending per employee at the heaviest US corporate users fell almost 10% in August as token prices dropped 41% since March and buyers shifted workloads to cheaper models.
The most aggressive corporate buyers of AI in the United States pulled back their per-employee spending by nearly 10 percent in August, according to the September 2026 edition of the Ramp AI Index, which tracks card and invoice data across American businesses. The drop landed hardest among the top 1 percent of companies by AI outlay, the cohort that had been bidding up frontier model contracts through the first half of the year. It marks the first clear contraction in that group's monthly run rate since the index began separating heavy adopters from the broader market.
The arithmetic behind the decline is straightforward: the price of a million tokens has fallen 41 percent since March 2026. Rather than simply spending less on the same workloads, companies are routing tasks that once went to the most expensive frontier systems toward cheaper alternatives, including smaller models from the same vendors and open-weight options. Ramp's data shows the shift is deliberate, not a slackening of demand — usage keeps climbing while the unit economics improve.
That distinction matters for OpenAI, Anthropic and the other labs selling at the top of the price ladder. If customers keep migrating work to lower-cost tiers, providers need volume to grow fast enough to offset the per-token decline. The index does not resolve whether that is happening; it only shows that the heaviest buyers have found ways to do more with a smaller budget line. For finance chiefs who spent 2025 defending six-figure AI invoices, August is the first month the spreadsheet moved the other way.
The pattern mirrors what usually happens when a technology moves from experimentation to production. Early adopters pay premium prices for capability, then optimize once workflows settle and competing suppliers emerge. The 41 percent slide in token pricing since March gives procurement teams real leverage for the first time, and the top 1 percent appear to be using it. Smaller companies, which never bought frontier capacity at scale, have less to gain from the same renegotiation.
What to watch is whether September and October reverse the dip or confirm it. If per-employee spending keeps falling while total token consumption rises, the labs face a margin squeeze that pricing pages alone cannot fix. If the top 1 percent resume spending at higher volumes, August will look like a one-month pause rather than a turning point. Ramp's next update, due in October, will carry the first real evidence either way.