The companies most focused on AI invest $7,500 per employee each month, while the median expenditure is $11.

      **Summary:** According to Ramp data, the top 1% of firms allocate $7,500 per employee each month on AI, while the top 10% spend $611, and the median is just $11.38. Notably, spending among power users saw an increase of 14.1% last month. The Ramp AI Index reveals a stark disparity in AI investments across American businesses.

      The "AI-pilled" top 1% are still not spending more on AI than on personnel, as U.S. software engineers earn about $16,000 monthly, significantly above the $7,500 AI expenditure. However, their investment trajectory is steep, with a 14.1% rise in AI spending per employee observed in the last month.

      The top 10% invest approximately $611 per employee monthly, which covers a few enterprise AI subscriptions and API usage. The median spending of $11.38 equates to essentially one subscription. For the majority of companies, AI expenditures remain minimal relative to their software budgets.

      This data coincides with anecdotal evidence from the high end of AI spending. An executive from Nvidia stated that compute costs now surpass employee salaries, while Mercor’s CEO mentioned that the startup's spending on tokens for internal projects exceeds its payroll. Despite a 98% reduction in per-token prices, enterprise AI costs have tripled, as these tools amplify consumption rates by 18.6 times per developer.

      There is a distinct paradox: while tokens are inexpensive, their usage is considerable. For instance, a simple linear workflow in 2023 costs around $0.04 per interaction, while a more complex system in 2026 will soar to approximately $1.20, making it about 30 times pricier. Microsoft has reported individual engineers spending between $500 to $2,000 monthly on Claude Code tokens, and Uber expended its entire AI coding budget for 2026 by April.

      Firms that are heavily invested in AI tend to utilize multiple advanced models and platforms, seeking out cheaper open-source options. This approach indicates that these power users are not loyal to any single provider, instead optimizing for both cost and functionality simultaneously.

      The critical question remains whether the $7,500 per employee figure represents a maximum limit or a foundation. If agentic AI continues to broaden the automation scope for companies, and if token expenses become a significant cost category alongside personnel and software, the leading 1% may resemble the median firm in a few years. Presently, however, most American companies spend less on AI than on a good lunch.