Companies are using more AI than ever. They just aren't paying more for it. That's the finding in a new Ramp report covered by TechCrunch: token spend across Ramp's customers grew only 1.5% over August and September, compared with 42% month over month in July, even as token volumes kept rising.
Ramp lead economist Ara Kharazian points to two forces. First, price competition. OpenAI cut API prices in half for GPT-6 Sol and GPT-6 Luna in September, and Anthropic followed with cuts to Fable 5.1 and Opus 5.5, a move the company said in a recent filing trimmed revenue growth by about $600 million, per TechCrunch. Second, buyers are getting thrifty: budgets per employee, defaults that block expensive models on long-running tasks, and routers that push work onto cheaper models.
"You can use more AI, but use it for less," Kharazian said. "That's great for the customers…but it's not necessarily great for the model companies."
The concentration is the risky part. "Almost the entire spend on AI is from venture-backed, high-growth tech companies," he said, and the top 1% of firms, which hover near $10,000 per employee, slipped slightly in August and September. Median firms spend $12-$15 per employee per month. October did re-accelerate, with token spend up 8.9% from September 23 to October 23, so one soft stretch isn't a trend.
Why it matters: The timing is awkward. Bloomberg reported Friday that Anthropic's potential valuation has fallen from about $2 trillion in early October to $1.2 to $1.3 trillion, and that OpenAI trimmed its fundraising goal, per TechCrunch. Frontier-lab valuations assume revenue keeps compounding. Ramp's data says price wars can turn rising usage into flat revenue, a buyer's market where discounts meant to lock in customers mostly cap growth.
The caveat: Ramp sees only its own customers, a startup-heavy slice, and Kharazian's thesis that coding agents were a one-off is still a view, not a fact.
What to watch: whether the October rebound holds in Ramp's next monthly index, and whether Anthropic's IPO filing shows the same squeeze in its own numbers.
— via TechCrunch