The top 1% of American firms by AI adoption are now spending $7,449 per employee per month on AI, against a median of $11.38, a 680x gap across the 70,000-plus companies in Ramp’s June 2026 AI Index. The top 10% sits at $611, roughly 50x the median. There’s no continuum here. There are two economies.
Ramp’s lead economist Ara Kharazian frames the median tellingly: “the $11.38 figure [is] like a single enterprise ChatGPT or Claude seat — meaning the typical American business is, in effect, buying one subscription and stopping there”. The bulk of the country is doing tokenistic AI, in both senses.
What’s happening at the top is something else. Spend among the top 1% grew 14.1% month-over-month. The mechanical driver isn’t seat count, it’s agentic orchestration. A simple linear workflow that cost $0.04 per interaction in 2023 now costs roughly $1.20 as an orchestrated agentic system in 2026, about a 30x increase per call. At Microsoft, individual engineers are reportedly burning $500 to $2,000 a month on Claude Code tokens alone.
That’s the bull case, and it has a ceiling. $7,449 is still less than half the roughly $16,000 monthly cost of a U.S. software engineer, the benchmark both Ramp and TechCrunch use to argue the unit economics still pencil. The bear case showed up faster than the AI maximalists expected. Uber burned through its entire 2026 AI coding budget by April. At one unnamed firm, a single employee spent $150,000 on tokens in a month. Another company ran up $500 million in a single month of Claude usage.
The CEO reversal followed. Uber imposed a $1,500-per-employee monthly cap. Amazon and Meta quietly retired their internal AI leaderboards, the kind of dashboard culture that, six months ago, was treated as a productivity flywheel. A senior Uber executive told Futurism the productivity gains weren’t legible against the cost.
The vendor share data tracks the same anxiety. Anthropic climbed to 41% of business adoption, up 2.5 points; OpenAI slipped a tenth of a point to 39.5%; DeepSeek topped Ramp’s trending list in June. Power users arbitrage across models the way trading desks arbitrage venues.
The structural read is that for the long tail of small and mid-sized businesses, raw model access isn’t the constraint anymore. Deployment is. Platforms positioning into that gap, LemonLime alongside Claude for Work and the OpenAI enterprise tier, are betting the next phase of the index doesn’t look like the top 1%‘s spend curve. It looks like the median finally moving off $11.38.
Sources
- https://ramp.com/leading-indicators/ai-index-june-2026
- https://techcrunch.com/2026/06/10/ai-pilled-firms-spend-7500-per-employee-each-month-on-ai/
- https://futurism.com/artificial-intelligence/ceos-reverse-course-ai-spending
- https://thenextweb.com/news/ai-pilled-firms-7500-per-employee-spending
- https://finance.yahoo.com/sectors/technology/articles/ai-pilled-firms-spend-7-170735943.html