Ramp AI Index: US business AI spending falls while usage rises about 50% from July peak

US businesses are spending less on AI while using more of it, according to the latest Ramp AI Index from Ramp economist Ara Kharazian. Usage has climbed about 50% since spending peaked in July and reached a record high at the end of September.

Key findings

• Spending is down from its July peak. Kharazian says this is not the first decline in the index, but it has lasted unusually long.
• Usage is up. Usage has risen roughly 50% since the July spending peak and hit a record at the end of September.
• Price competition is the cause. Kharazian points to price cuts on top models and to cheaper, more efficient standard and lite models. He says the drop comes almost entirely from competition between OpenAI and Anthropic.
• Anthropic leads on token spending. In the last week of September covered by the data, Anthropic took 51% of token spending and OpenAI took 44.5%. Together the two account for 95.5%, which leaves 4.5% for all other providers.
• Open-source models remain marginal. Kharazian says they still make up less than 5% of business spending.

Methodology and limits

The Ramp AI Index tracks transaction data from more than 70,000 US companies. The token data comes from a subsample of that group. Kharazian says the analysis covers only API spending and skews toward large AI customers. Spending through consumer subscriptions or seat-based enterprise plans is not captured in the token figures, and the results may not represent smaller businesses.

The source does not give dollar figures for total spending, the size of the decline, or the specific price cuts and model releases behind it. Those details are not disclosed in the available reporting.

What this means

The data points to falling unit prices, not falling demand. If usage is up about 50% while spend is down, the effective price per token paid by these companies has dropped by more than the headline usage figure alone suggests. That fits Kharazian's explanation: price cuts on flagship models plus cheaper standard and lite tiers.

The market is close to a duopoly among API buyers in this sample. Anthropic and OpenAI together hold 95.5% of token spend, and open-source models are under 5%. A price war between two vendors can cut customer bills quickly, but it can also reverse. Buyers with large API commitments are benefiting now, and the sample's skew toward large customers means the effect is probably strongest for them.

For builders, the practical effect is that workloads that were too expensive at July prices may now be viable, particularly on the lighter model tiers. The caveat is that this is one transaction-based index with a token subsample and API-only coverage. It shows direction and rough market share, not a full picture of enterprise AI spending.