Recent data shows a dip in AI budget per head at some of the biggest spenders during August, raising questions about whether this is a seasonal lull or a deeper shift in business behavior. Payments provider Ramp surveyed over 70,000 companies and found that 56% paid for AI products in August—only up slightly, by 0.4%, from July. That modest rise continues a pattern of flattening after earlier surges in AI adoption and expenditure.
Top Tier Pulls Back
The most striking change came from the top 1% of AI-spending firms, where spending per employee fell nearly 10% to about $7,205. Meanwhile, costs per million tokens dropped sharply, from a March 2026 peak of $1.15 to $0.68. That reflects not only tougher price competition—especially between OpenAI and Anthropic—but also a move by companies to opt for older or cheaper models rather than premium frontier releases.
Smaller and mid-tier companies continue to show little AI spend depth. For example, the median AI-buying firm still spends only around $11.95 per employee. The top 10% spend about $650 per head—vastly less than the heavyweight 1% tier. These numbers underscore the extreme disparity in how deeply different businesses are investing in AI.
Is It Just the August Effect?
August is often a slow month for business, and Ramp’s economist suggests some of the drop may be explained by vacations and lower usage. However, there are warning signs that signal this may go beyond a seasonal dip. Lower token costs and shifting usage toward cheaper models could erode the profitability of expensive frontier model launches. Heavier spening firms may struggle to justify the upfront investments if adoption slows.
Another data point from the U.S. Census Bureau paints a different picture: only about 22% of all businesses report using AI, suggesting the Ramp sample skews toward more tech-forward companies. What this means is the broader market may lag even further in spending.
For now, there’s also a strategic shift among AI labs and vendors. With lower cost per token, businesses are choosing older, lower-priced models like ChatGPT 5.6-Terra and Anthropic’s Sonnet instead of premium frontier ones. Only around 6.4% of AI-spending companies are using inference or model-serving platforms in August—a growing segment, but still far from central to mainstream adoption.
Still, companies investing heavily in AI are adding headcount, not cutting it—an indication that for those doubling down, AI is expanding work, not replacing it.
This pullback may be a temporary blip or the beginning of a larger calibration. For hyperscalers and model-builders with large infrastructure spending, slower revenue growth could challenge margins. The trend toward lower-cost, lower-usage models suggests early excitement has given way to more cautious investment. Whether August becomes a turning point or just another seasonal dip depends heavily on whether usage and value keep up for the companies spending the big money.
