Corporate AI spending in the U.S. is becoming an increasingly lopsided game. The top 1% of American businesses spent a record median of $7,400 per employee per month on AI tools in July, according to Ramp's AI Index, which tracks adoption through the corporate card and expense-management company's own transaction data. That compares with just $650 per employee for the top 10% of firms and a median of $11.95 across all companies Ramp tracks — a roughly 680-fold gap between the most AI-intensive firms and the typical business.
The concentration has been building for some time. As recently as early 2024, even the top 1% of firms spent below $1,000 per employee per month on AI tools. Spending among that leading cohort has more than tripled over the past several months alone, growing 14.1% per employee in July compared to the prior month — an unusually fast pace for a metric already at record highs.
A widening gap, not a rising tide
Related: Sandisk On-Chain Perpetual Open Interest Hits Record $1.73B
The scale of the divide suggests AI adoption isn't spreading evenly across corporate America so much as concentrating hard among a small group of firms already committed to heavy usage. Ramp's data shows top-percentile companies tend to spread spending across multiple frontier AI models and platforms rather than standardizing on a single vendor, often mixing premium frontier models with cheaper open-source alternatives depending on the task — a strategy that itself requires more sophisticated internal tooling and know-how than simply signing up for one AI subscription.
Even at $7,400 per employee monthly, AI spending among the most aggressive adopters still sits well below typical software engineer compensation, which averages roughly $16,000 per month in total pay. But the trajectory matters as much as the current level: an Nvidia executive was recently quoted noting that for some AI-heavy firms, the cost of compute has already overtaken the cost of employees — a threshold that, if the current growth rate holds, more companies are likely to cross before the concentration in AI spending starts to broaden out rather than deepen further.
The AI cost curve is bending both ways at once
The spending concentration data offers a useful counterpoint to a separate trend: even as AI usage costs are climbing for the heaviest corporate adopters, the underlying per-token price of AI model access has been falling sharply as providers compete on cost. That combination — falling unit prices alongside rising total spend among top users — points to expanding usage volume, not rising per-query costs, as the real driver behind the top 1%'s ballooning AI budgets.
