What can we learn from Vercel data?

I want to ask the same questions we asked with OpenRouter. The problem is that Vercel does not publish the exact number of tokens or dollars. It publishes growth and market share. So we can see how fast it is growing, but not how big it is.

Token volume

April 2026 is set to 100. This shows relative growth, not the number of tokens.

May grew 20% from April. June grew another 29%.

How fast is token use growing?

Change from the previous month.

Estimated spending

April 2026 is set to 100. Vercel values BYOK traffic at the providers’ published prices, so this is market-rate spend, not confirmed Vercel revenue.

Spending grew faster than tokens across these two months.

Estimated price per token

Spending index divided by token index. April 2026 is set to 100.

The average token became about 19% more expensive in May and stayed roughly flat in June.

How fast is estimated spending growing?

Change from the previous month.

Who is getting the traffic?

Daily share of all Vercel AI Gateway traffic in the public export.

Measured by

What we cannot see: absolute token volume, absolute spending, Vercel revenue or the value of one index point. The daily leaderboard also begins on June 19, 2026. The monthly index has only three observations, so it is not enough to establish a long-term growth rate.

Sources: Vercel’s May, June and July Production Index reports, and its public leaderboard export, licensed CC BY 4.0.