Businesses are using more artificial intelligence than ever, but they have stopped spending more on it. New data from Ramp, the corporate payments platform, shows spending on AI tokens grew just 1.5% across August and September, after surging 42% in July. Token volumes kept rising through the stall, which points to a price war rather than a demand problem: the frontier labs are cutting prices so aggressively to win customers that revenue growth has flattened even as usage climbs.
The stall in the numbers
Ramp tracks spending by more than 85,000 businesses, and its token figures come from customers using its Token Spend Management product. The company cautions that the token sample is not representative of its total AI spending, though it says the results run in the same direction. The shape is striking: two straight months of no growth after a summer surge.
The slowdown is concentrated at the top. Ramp says 1% of firms drive 80% of spending on OpenAI and Anthropic, and that cohort has stuttered: the heaviest users spent just below $10,000 per employee a month in August and September, down from their summer peaks.
Growth further down the ladder cannot make up the gap. Median firms spend only $12 to $15 per employee a month on AI, and the top 10% spend $600 to $1,000. Ramp’s lead economist, Ara Kharazian, told TechCrunch the top 1% matters so much that its wobble alone is enough to slow revenue growth entirely, despite rising volumes.
A price war, not a demand slump
Kharazian puts the cause squarely on competition, not demand. In September, OpenAI halved API prices for its GPT-6 Sol and GPT-6 Luna models, and Anthropic answered with cuts of its own for Fable 5.1 and Opus 5.5. Anthropic said in a recent filing that those cuts and contractual discounts reduced its revenue growth by about $600 million.
The price collapse shows up in Ramp’s index. The blended price per million tokens for OpenAI and Anthropic fell 26.8% in the four weeks to October 23, from $0.76 to $0.56, while token volume rose 37.1%. Anthropic’s share of token spending climbed from 51.3% to 60.5% over the same weeks, while OpenAI’s fell from 44.2% to 35.1%.
October brought a partial rebound, with token spending up 8.9% from September 23 to October 23. But Kharazian says the headwinds on spending growth are strong, and he told TechCrunch the “era of hypergrowth may be over.” The stall lands just as Anthropic approaches its public listing, with investors already questioning the valuations attached to the leading labs.
Our Read
The winners here are the buyers. Businesses get more AI for less money, and they are getting better at routing work to cheaper models, imposing budgets and skipping the most expensive systems for routine tasks, as they learn what each job actually needs.
It is a harder market for the labs, which are subsidising usage today to lock in customers they hope to charge more later. That playbook worked for ride-hailing, but Kharazian argues it will not work here: with Meta, Nvidia and open-model builders all offering cheap alternatives, there is too much competition for prices to rise again later.
There is a second squeeze coming from inside the labs. Independent measurement found Anthropic improving about twice as fast as a year ago, largely because its own AI does research work — and faster research means cheaper models sooner for everyone.
If that compounding continues, the price floor keeps dropping, and the labs’ bet has to be that new uses, not higher prices, will pay for it. Kharazian is sceptical that personal agents will be the next driver: they are, in his view, not as economically productive as coding agents, which powered the last growth wave.
Why It Matters
Revenue trajectories are the foundation of the trillion-dollar valuations now attached to the leading labs. If businesses keep using more AI while spending barely grows, the labs must either find new sources of revenue or accept thinner economics — and the public markets are already repricing the trade, with AI stocks sliding sharply in the past week.
Background
Ramp first flagged the flat August and September in its October 14 AI Index update, titled “Cracks in the AI Thesis.” Bloomberg reported on Friday that Anthropic’s potential valuation had fallen to $1.2 trillion to $1.3 trillion from about $2 trillion in early October, while OpenAI cut its proposed valuation to $1.25 trillion from $1.4 trillion.
Source: Ramp’s AI Index and TechCrunch’s report
0 Comments