Business of GTM

Ramp Says Business AI Token Volume Climbed 37.1% While Spend Rose 8.9%

Ramp's AI Index shows businesses pushed far more work through AI models in a month while paying only a little more, as OpenAI and Anthropic cut prices to win share.

Ramp CEO Eric Glyman seated on stage and gesturing, in black and white over a red block on a beige background.
Ramp CEO Eric Glyman at a Citadel Securities fireside conversation. Photo illustration by Business of GTM. Original image: Citadel Securities.

TL;DR

  • Ramp's AI Index shows business token volume up 37.1% in a month while token spend rose only 8.9%.
  • We think the frontier labs are buying share with price cuts, which means usage growth no longer guarantees revenue growth for consumption-priced sellers.
  • Anthropic's next filing and Ramp's next monthly index will show whether volume keeps outgrowing the discounts.

Businesses are pushing far more work through AI models and paying only a little more for it.

Ramp's AI Index is monthly. Its Tokenomics section compares the latest complete week, ending October 23, with the same week a month earlier, ending September 25. Token volume rose 37.1% while token spend rose 8.9%.

Over the same two weeks, the blended OpenAI and Anthropic price per million tokens fell 26.8%, from $0.76 to $0.56. TechCrunch tied the data to Anthropic's coming IPO, saying it suggests investors may be wrong to assume breakneck growth.

Flat spend can look like fading demand, but usage kept climbing. The price fell instead, as the two leading labs cut it to win share. For a B2B company that sells on consumption, that pulls apart two numbers that usually move together: usage and revenue.

Anthropic is winning the share fight

In its October write-up, Ramp lead economist Ara Kharazian said token spend stayed virtually flat through August and September. Early-October spend was only 1.5% above the August peak. Kharazian puts that down to frontier price cuts and a falling blended price per token.

The weekly data shows who gained. Anthropic's share of token spend rose from 51.3% to 60.5% between the two weeks, and OpenAI's fell from 44.2% to 35.1%. OpenAI cut harder. Its blended price fell 42.4%, against 15.7% for Anthropic.

Kharazian also reports that enterprise contracts come with discounts and that "the customer is in the favorable negotiating position." That comes from what businesses told Kharazian, and Ramp's spend data doesn't measure it. TechCrunch reports customers adding spending caps, cheaper default models and routers.

Anthropic's own numbers show what that position costs the seller. Its amended S-1 puts Q3 revenue at $17 billion, up about 47% from $11.6 billion in Q2, Fortune reports. Growth was about 144% the quarter before. About $600 million of the Q3 gain was offset by "price reductions introduced as part of new model versions and contractual discounts." Adjusted operating margin fell to 15% from 19%.

Growth depends on usage outrunning the cuts

Ramp's data and Anthropic's don't contradict each other. Ramp tracks spend by US businesses on its platform. Anthropic reports all of its revenue for a calendar quarter. A 47% quarter can sit beside nearly flat business spend on Ramp because the two count different buyers over different periods.

Both point to the same mechanism. Ramp's October data already put two labs well ahead in business adoption, Anthropic at 43.8% and OpenAI at 40.6%, and those two now compete on price. Anthropic's revenue grows only while usage grows faster than its price falls, and each new model version resets the price lower. Once the cuts outpace usage, the usage chart stops predicting the revenue line. The same holds for any seller billing on usage to buyers who now set spend caps and route to the cheapest model.

Kharazian's own warning is about who is buying. He wrote on X that the top 1% of companies drive 80% of AI spend, and 80% of those are venture-backed. The risk he names is "a cascading effect of large customers simultaneously pulling back their AI spend." That would hit volume and price at the same time.

The weekly token figures come from businesses using Ramp's Token Spend Management product. Ramp says that sample isn't representative of its total AI spend, though "directionally similar." Ramp sells AI spend controls, so cost discipline is a story it profits from. Anthropic's figures are company-reported in a filing.

The discount line in Anthropic's next filing, set against the $600 million that price cuts and contractual discounts took out of Q3, will show whether volume is still outrunning the cuts.

Usage is now the easy number, and revenue per unit decides

Ramp’s figures cover two AI labs, but any company that bills on usage now sells to buyers with spend caps, cheaper default models and routers that pick the lowest price. In that market, usage and revenue per unit can move in opposite directions, and a plan built on the usage chart alone will overstate revenue. Anthropic’s filing shows both at once, with volume up and $600 million of one quarter’s gain given back to price.

the customer is in the favorable negotiating position.

Ara Kharazian, lead economist at Ramp

Anthropic showed both numbers in the same filing. Volume up, and $600 million of a quarter's gain given back to price.

By Business of GTM