Ramp Data Shows AI Spending Stalling as Anthropic Nears IPO

Ramp’s corporate card data shows AI spending flattening in late summer even as token volumes climb, raising doubts about the hypergrowth assumptions underpinning Anthropic’s planned IPO and OpenAI’s fundraising. Across August and September, AI token spend grew just 1.5%, a sharp slowdown from July’s 42% month-over-month increase, despite rising usage. October data briefly reaccelerated, with token spend up 8.9% from September 23 to October 23, but Ramp lead economist Ara Kharazian said structural headwinds remain strong. The spending slowdown appears concentrated among the most intensive AI users. Kharazian said the top 1% of firms on Ramp’s platform, which had been spending near $10,000 per employee per month during the summer, dipped slightly in August and September. Growth from median firms, which spend $12 to $15 per employee per month, and the top 10%, which spend $600 to $1,000, was not enough to offset the decline at the top. He said many use cases outside software and finance may never support spending at the levels seen among the most advanced buyers. Ramp attributes the trend less to weakening demand than to aggressive pricing competition and more disciplined AI usage. OpenAI cut API prices by half for GPT-6 Sol and GPT-6 Luna in September, and Anthropic lowered prices for Fable 5.1 and Opus 5.5, a move the company said reduced revenue growth by about $600 million. Enterprise discounts aimed at locking in customers are also compressing revenue. At the same time, customers are using cheaper models more efficiently, setting default usage limits, imposing employee budgets and routing workloads to lower-cost options. Tools such as Jev, which helps select among predetermined outputs, further reduce costs for simple tasks. Kharazian described the market as increasingly buyer-oriented, arguing that Big Tech’s historical playbook of subsidizing technology to win customers may not work in AI because competition is too fragmented, including open models from China and the United States, enterprise offerings from Meta and Nvidia, and newer lightweight products. Notion co-founder Akshay Kothari pointed to internal data showing token usage rising while costs fall, reinforcing the pattern. The findings have arrived as investors reassess frontier AI companies. Bloomberg reported Anthropic’s expected valuation had fallen from about $2 trillion in early October to $1.2 trillion to $1.3 trillion, with expectations of a high junk rating from credit agencies. OpenAI also reportedly cut its proposed valuation to $1.25 trillion from $1.4 trillion and reduced its latest fundraising goal by about $10 billion. While some investors hope new products such as personal agents could renew spending growth, Kharazian is skeptical, noting they are less economically productive than coding tools. For now, Ramp’s data suggests companies can consume more AI while paying less, a benefit for customers but a challenge for model providers betting on sustained revenue acceleration.