Artificial intelligence is getting more expensive inside companies, but not expensive enough to overtake people. New data from Ramp’s AI Index suggests that even the most AI-heavy businesses are still spending more on employee labor than on AI services and compute, despite loud claims from tech leaders that the balance has already flipped.
Ramp’s AI Index puts monthly AI spend at about $7,500 per employee in the top 1% of companies by AI adoption. By comparison, the average software engineer salary in the US is about $16,000 a month, leaving payroll well ahead of AI costs for now.
Ramp’s AI Index shows a smaller bill than the headlines
That matters because recent comments from Nvidia and the startup Mercor helped fuel a simple but dramatic narrative: AI compute is already more costly than human work. Ramp’s data says the picture is less theatrical. The median company in the index spends just $11.38 per employee on AI each month – basically the price of a basic business AI subscription – while the top 10% average about $611 per employee.
There is still momentum, though. Ramp says compute spending per employee among the most AI-active companies rose about 14.1% over the last month. That is a quick climb, but it is also a reminder that this market is still noisy, with costs bouncing around as companies test tools rather than locking into a single AI stack.
Hybrid model shopping is keeping costs in check
The biggest clue is how companies are buying AI. The most aggressive adopters are not betting on one model or one vendor; they are switching between large language models and mixing premium proprietary systems with cheaper open-source options. That kind of shopping behavior is boring, but effective. It stops AI bills from ballooning as fast as the product demos suggest.
That also puts the current wave in a broader business context: cloud software adoption took years to become a major budget item, and AI may follow a similar path. The spend is real, the growth is real, but the idea that most companies are already paying more for tokens than for talent is not supported by this dataset.
AI spending growth is still climbing
The more interesting question is how quickly those costs become structural. If AI usage keeps rising at double-digit rates while companies keep blending models to chase savings, the bill will grow – just not as neatly or as fast as the loudest predictions claim. For now, payroll still sits at the top of the expense chart, and AI is the annoying new line item climbing up behind it.