THE ESSENTIALS IN 30 SECONDS

  • 55.9% of US businesses on Ramp paid for AI in September, up 0.6 points; Anthropic was used by 43.8% and OpenAI by 40.6%.
  • Ramp says token spending saw virtually no growth in August and September even as volumes rose; per-employee spend at the top 1% of firms slipped 3.1% from its July high.
  • Ramp's lead economist attributes the slowdown to price competition and cost control rather than weaker demand; TechCrunch reports token spend rose 8.9% in the month to 23 October.

Spending data from Ramp, the US corporate card and expense company, suggests that business outlays on AI models largely stalled in late summer after a sharp jump in July. TechCrunch highlighted the figures on 6 November, noting that they arrive as Anthropic, one of the two leading suppliers in Ramp's data, prepares its stock market debut.

What the numbers show

Ramp's AI Index, based on card, invoice and bank transfers from more than 85,000 US businesses, found that 55.9% of them paid for an AI product in September, up 0.6 percentage points from August. Anthropic was paid by 43.8% of businesses and OpenAI by 40.6%, with OpenAI narrowing the gap. Ramp notes that free tools and personal accounts are not captured, so real usage is likely higher.

Spending tells a different story. In the October edition of the index, lead economist Ara Kharazian writes that token spending showed virtually no growth in August and September while the number of tokens consumed kept rising. TechCrunch puts the two-month increase at 1.5%, after a 42% jump in July. Among the heaviest users, the top 1% of firms, monthly AI spend per employee fell to $9,568 in September, 3.1% below a July peak of $9,873. Median firms spend roughly $12 to $15 per employee per month, according to TechCrunch. Token data comes from a separate sample of Ramp customers and is not representative of all AI spending, the company cautions.

Why it matters

Kharazian argues the slowdown reflects falling prices and tighter cost control rather than weaker demand. Frontier labs have cut list prices and, based on his conversations with customers, offer discounts in contracts that rarely assume unusually high usage. He also suggests adoption may be nearing the top of an S-curve, so future growth would depend on existing customers using more. He told TechCrunch that nearly all AI spending comes from venture-backed tech firms, which leaves the market exposed if they all cut back at once.

TechCrunch reports that token spend re-accelerated by 8.9% between 23 September and 23 October, so the trend is not one-directional. It also cites an Anthropic filing saying price cuts on some of its models reduced revenue growth by about $600 million. Ramp stresses that its top-1% estimates are volatile and that its results may be revised.