The key facts
- Token spending in August–September grew by just 1.5% after a 42% rise in July (Ramp)
- Price cuts by OpenAI and Anthropic cost Anthropic about $600 million in revenue growth, according to its statement in filings
- Anthropic's IPO valuation fell from $2 trillion to $1.2–1.3 trillion, OpenAI's from $1.4 trillion to $1.25 trillion
Spending stalls while usage grows
Companies are using AI more than ever, but are not spending more on it. According to a report from the Ramp platform, AI token spending grew by just 1.5% in August and September, whereas in July the increase was 42% compared to the previous month. Token consumption volumes, meanwhile, kept growing — meaning the issue is not falling demand, but falling prices. TechCrunch reports this citing Ramp data.
A price war and cost-conscious customers
According to Ramp's chief economist Ara Khachaturyan, the AI market has effectively become a buyer's market: in September, OpenAI cut API prices for the GPT-6 Sol and GPT-6 Luna models in half, and Anthropic responded by significantly lowering prices for Fable 5.1 and Opus 5.5 — according to the company's statement in recent filings, this reduced revenue growth by roughly $600 million. Both companies also offer enterprise customers discounts to lock in spending. An additional factor is that customers themselves have learned to economize: they cap employee budgets, route some tasks to cheaper models via routers, and use new lightweight tools such as the company Jev, which picks between several pre-set answers and already ranks among the best performers on Ramp's metrics. In October, spending accelerated again — up 8.9% from September 23 to October 23 — but Khachaturyan believes the headwinds are strong.
Dependence on the top 1% of customers
The key vulnerability, according to Ramp, is that nearly all AI spending comes from fast-growing venture-backed tech companies. The top 1% of the most 'AI-dependent' firms spent up to $10,000 per employee per month at the summer peak, but in August and September their spending slowed slightly. The median company spends only $12–15 per employee per month, the top 10% spend $600–1,000, and their growth does not offset the slowdown in the top segment. Khachaturyan doubts that ordinary businesses — medical practices or car dealerships — will ever spend comparable amounts. A similar picture was published by Notion co-founder Akshay Kothari based on the company's internal data: 'tokens are growing, spending — ouch.'
Valuations fall ahead of the IPO
The slowdown coincided with Anthropic's preparations for its IPO. According to Bloomberg, its potential valuation dropped from the $2 trillion expected in early October to $1.2–1.3 trillion; in addition, agencies are reportedly preparing to assign its bonds a high 'junk' rating. OpenAI also lowered its target valuation from $1.4 trillion to $1.25 trillion and cut the goals of its latest funding round by roughly $10 billion. Khachaturyan believes Ramp's data is largely fueling investor concerns. The industry's bet is on new product categories such as personal agents, which have already been released by OpenAI, SpaceXAI and Meta along with the startup Instinct — but he is skeptical: personal agents are 'not as economically productive as coding,' and the observable products so far do not offset the spending slowdown.



