Corporate AI use keeps climbing. Budgets stopped.
Token spending on Ramp’s corporate card rose 1.5% across August and September, the fintech revealed. In July that figure jumped 42% month over month. October showed a partial rebound at 8.9%.
The cause is visible in pricing. OpenAI took its September rates for GPT-6 Sol and GPT-6 Luna down by half, while Anthropic told investors its cuts on Fable 5.1 and Opus 5.5 trimmed roughly $600M from revenue growth.
The top 1% stumbled first
Ramp economist Ara Kharazian points at the heaviest buyers. The most AI-committed firms, which had been spending about $10,000 per employee each month at their summer peak, slipped below that mark in the same two months. Because a small group drives so large a share of spend, their hesitation moved the whole index.
Everyone else is far behind anyway. Median companies spend $12 to $15 per employee monthly. The top decile runs $600 to $1,000. Even with faster mid-market adoption, four figures per head looks implausible for a dentist’s office or a car dealership.
Cheaper routing eats the difference
Buyers grew thriftier. Teams cap expensive models for long tasks, set per-employee budgets, and route simple requests through small models. Kharazian cites Jev, a tool that picks among preset answers, as a fast-growing line item. For many jobs, a yes or no is the entire requirement.
The timing is awkward. Bloomberg reported Friday that Anthropic’s expected IPO valuation slid from about $2T to $1.2T to $1.3T, and that credit agencies may hand it a high junk rating. OpenAI reportedly trimmed its own target to $1.25T while cutting roughly $10B from its latest raise.
Notion co-founder Akshay Kothari summed up the mood in a line that traveled quickly: tokens go brrr, costs go ahhh.
