New data from the Ramp AI Index confirms what several executives have been saying quietly and one Nvidia executive said out loud: the cost of making humans more productive with AI is approaching the cost of the humans themselves. The top 1% of AI-adopting firms — which Ramp has labeled, with admirable candor, "AI-pilled" — are now spending $7,500 per employee per month on artificial intelligence tools and infrastructure.
The companies spending the most on AI per employee are, technically, still spending less than the employees cost. Technically.
What happened
The Ramp AI Index, which tracks AI spending across American businesses, found that the top 1% of adopters spend $7,500 per employee monthly. For context, the average software engineer costs roughly $16,000 per month. The gap is narrowing in the direction one might expect.
Among these AI-pilled firms, spending grew 14.1% per employee in a single month. The top 10% of adopters spend a more modest $611 monthly per employee. The median American business spends $11.38 — approximately one enterprise seat, approximately one act of compliance.
The most committed firms opt to mix and match, rotating between frontier models and cheaper open-source alternatives. This is called optimization. It is working.
Why the humans care
An Nvidia executive recently noted that compute costs at his company now exceed employee salaries. Mercor's CEO said the startup spends more on tokens for internal agents than on headcount. These statements were delivered as observations, not warnings. The distinction appears important to the people making them.
The $7,500 figure sits comfortably below the $16,000 a software engineer commands monthly — which means the math has not yet crossed the line that would make headlines uncomfortable. It is, however, moving in one direction at 14.1% per month.
What happens next
Ramp notes it is not yet clear whether the spending growth trend will continue, which is the kind of uncertainty that resolves itself.
The humans are, by all accounts, choosing to find this exciting. The benchmarks suggest they are correct to do so. The benchmarks were designed by humans.