The most aggressive corporate adopters of artificial intelligence are pouring roughly $7,500 per employee per month into AI tools, according to new data from the Ramp AI Index, which tracks adoption across American businesses. That figure refers to what Ramp calls the "AI-pilled" top 1% of firms, a group whose spending has surged 14.1% per employee in just the last month. To put that in perspective, the average software engineer in the U.S. earns about $16,000 a month, meaning even the heaviest AI spenders haven't yet eclipsed human salary costs, contrary to recent comments from industry insiders. An Nvidia executive recently remarked that compute now costs more than his employees' salaries, and Mercor CEO Brendan Foody said last week that his startup is spending more on tokens powering internal agents than on personnel. But the headline number masks how uneven AI adoption really is. The next tier of power users, the top 10% of companies, spend about $611 per employee each month, while the median firm shells out just $11.38, roughly the cost of a single enterprise seat on a SaaS platform. That gulf suggests that while a small group of companies has gone all-in on AI infrastructure, the bulk of American businesses are still in experimentation mode, treating the technology more like a budget subscription than a strategic line item. One reason the heaviest spenders can sustain those costs is that they don't lock themselves into a single provider. Ramp's data shows that the top 1% routinely mix and match between multiple frontier models, often pairing commercial systems with cheaper open-source alternatives to keep token costs manageable. Whether the 14.1% monthly growth rate among these power users continues is an open question, particularly as enterprises reassess budgets and pressure builds to show returns on AI investments. For now, Ramp's research suggests the AI spending explosion is real, but it's concentrated in a thin slice of the corporate world rather than a broad-based shift.