Ramp Data Shows AI Token Spending Flatlining Just as Anthropic Nears Its IPO

Companies are using more AI than ever, but they are no longer paying meaningfully more for it. New data from corporate spend platform Ramp shows that AI token spending grew only 1.5% across August and September combined, a stark contrast to July, when spend jumped 42% month over month. The stall comes even as token volumes, a pure measure of usage, continued to rise — a divergence that suggests price competition and savvier customers, not weakening demand, are capping revenue growth at OpenAI and Anthropic.

The timing matters. Anthropic is preparing for one of the most anticipated IPOs of the year, and investors have been assuming AI revenue keeps growing at breakneck speed. Ramp's numbers argue otherwise. According to Bloomberg, Anthropic's potential valuation has sunk from an estimated $2 trillion in early October to between $1.2 trillion and $1.3 trillion. OpenAI, separately, lowered its proposed valuation to $1.25 trillion from $1.4 trillion and cut roughly $10 billion from its latest fundraising goals.

"Almost the entire spend on AI is from venture-backed, high-growth tech companies," Ramp's lead economist Ara Kharazian told TechCrunch. "That's fine as long as AI spend continues to rise, but if it starts to slow down because of something in the market or some other factor, they will all start to slow down simultaneously, and that could lead to other cascading effects."

The slowdown is concentrated at the very top. The top 1% of firms — the most "AI-pilled" companies, which were spending close to $10,000 per employee per month at their summer peaks — stuttered in August and September, coming in just below that mark. Because that cohort accounts for so much of total spend, its pullback alone is "sufficient to slow down revenue entirely, despite the fact that we're seeing rising volumes," Kharazian said. By comparison, median firms spend $12 to $15 per employee per month and the top 10% spend $600 to $1,000, and their growth is not enough to make up the difference. He doubts a typical doctor's office or car dealership will ever approach the intensity of the top 1%.

Kharazian, who has published a blog series called "Cracks in the AI Thesis," attributes the stall to competition rather than demand. In September, OpenAI cut API prices in half for GPT-6 Sol and GPT-6 Luna. Anthropic responded with significant reductions for Fable 5.1 and Opus 5.5 — a decision the company said in a recent filing slashed its revenue growth by about $600 million. Both companies are also offering enterprise discounts meant to lock in spend, which in practice hit their revenue targets. "It very much has become a buyer's market," Kharazian said.

Cheaper alternatives are compounding the pressure. Meta and Nvidia are pushing enterprise offerings built on cheap or open-source models, open-weight labs in China and the US keep pricing down, and Jev — the lightweight decision tool that routes simple yes-or-no workloads to tiny models — is increasingly among the top performers on Ramp's platform. "You take your already cheap models, you shift them over to Jev, and now your already cheap stuff gets even cheaper," Kharazian said. Companies are also getting thriftier on their own: defaulting long-running tasks away from expensive models, imposing employee budgets, and shifting work through routers to lighter tiers. Notion co-founder Akshay Kothari summed up the pattern with internal data of his own: "tokens go brr. Costs go ahhh."

There is one flicker of recovery. October data showed a re-acceleration, with token spend up 8.9% from September 23 to October 23, though Kharazian said headwinds remain strong. The bigger open question is what the next growth engine would even be. AI coding emerged a year and a half ago as the commercially dominant use case and helped make Anthropic and OpenAI trillion-dollar companies; personal agents from OpenAI, SpaceXAI, Meta and startups like Instinct are the presumed successor, but Kharazian is skeptical they are "as economically productive as coding."

For the IPO market, the message is uncomfortable: usage is compounding, revenue is not. If the "era of hypergrowth" is over, as Kharazian put it, model companies will have to defend valuations priced for endless expansion with a product mix whose prices only move in one direction. The data is Ramp's own index of corporate card and bill-pay transactions and has not been independently audited, but it aligns with the valuation resets now rippling through both private markets and the public AI trade.