Vercel AI Gateway: AI expenditures multiplied by 13 in nine months, while the price of token fell by 47%

Vercel's October barometer shows AI budgets that explode despite falling prices, major open models in volume and one in two teams that changed their main model in September.

Vercel has published the October edition of its "AI Gateway Production Index", derived from the traffic that passes through its model access gateway. Between December 2025 and September 2026, its customers' spending jumped by 1,180%, or about thirteen times the starting level, while the average price of the token fell by 47% over the same period. The median active paying customer over the two months more than doubled their bill.

Open models dominate in volume

Open-weight models now account for 71.5% of tokens traded and 23% of spending; their share of the budget has more than quadrupled in three months, from one in twenty dollars in June to nearly one in four in September. More than two-thirds of spending on Chinese-sourced models goes through U.S. hosting providers, up from less than 10% in December.

Loyalty is low: nearly half of the teams changed the most used model in September and almost one in three changed labs, a record. In terms of launches, TypeSafe AI's Jev was adopted by 35.3% of paid teams in two weeks but only accounts for 0.27% of spend, Anthropic's Opus 5.5 captured 11.03% of the budget during its launch window, and OpenAI's GPT-6 Luna has the best retention, at 64.8%. Vercel points out that its sample, composed mainly of developers, is not representative of the entire market.