Ramp: AI spending growth stalled as Anthropic nears IPO

New data from expense-management firm Ramp shows AI spending barely grew over the summer even as companies used more of it than ever — a sign, Ramp’s lead economist Ara Kharazian told TechCrunch, that price competition and savvier customers may be capping revenue for OpenAI and Anthropic. Token spend rose just 1.5% across August and September, Ramp found, after climbing 42% in July alone; October brought a partial rebound, with spend up 8.9% in the month to October 23.

The slowdown is concentrated at the top. Kharazian said the most AI-heavy 1% of firms hit roughly $10,000 per employee per month at their summer peak, then slipped just below that level in August and September — enough on its own to flatten total revenue growth even as usage volumes kept rising. Median firms spend only $12–$15 per employee monthly; the top 10% spend $600–$1,000. Kharazian said it isn’t clear many AI use cases will ever justify the $10,000-per-employee mark the top 1% has reached — a doctor’s office or car dealership has little obvious reason to spend that much.

Kharazian pointed to the competitive market, not weak demand, as the driver: in September, OpenAI responded by halving its own API prices for GPT-6 Sol and GPT-6 Luna, and Anthropic followed with significant cuts to Fable 5.1 and Opus 5.5 pricing that it said in a recent filing slashed its own revenue growth by about $600 million. Enterprise discounts meant to lock in customers are, in practice, undercutting the companies’ own revenue targets — a dynamic Kharazian says breaks down because, unlike past subsidized markets such as ride-hailing, AI faces far more competition, including cheap or open models from Meta, Nvidia and Chinese labs, plus simple decision-tools like Jev that substitute for costlier models on basic yes/no tasks. Notion co-founder Akshay Kothari recently summed up the same dynamic: “tokens go brr. Costs go ahhh.”

The timing is awkward for both leading labs. Bloomberg reported Friday that Anthropic’s potential valuation has fallen to $1.2–$1.3 trillion, down from an estimated $2 trillion in early October, with the company also reportedly expected to draw a high “junk” credit rating; OpenAI has cut its own proposed valuation to $1.25 trillion from $1.4 trillion and trimmed its latest fundraising target by about $10 billion. Kharazian said a possible bright spot — new use cases like personal AI agents — is unproven, since such agents are less economically productive than coding tools have been.

Why it matters

If AI spending keeps growing only as fast as companies get thriftier about using it, the hypergrowth story underpinning trillion-dollar valuations for OpenAI and Anthropic gets harder to sustain.