Ramp, the corporate card and bill payment company, checked its records from more than 70,000 businesses and found something worth watching: the share of companies paying for AI tools grew by just 0.4 percentage points in August, reaching 56 percent. That is one of the slowest months of growth Ramp has recorded since it started tracking this.
Before anyone panics, this has happened before. Last year, Ramp's data showed nearly the same pattern: growth stalling out between August and October, then picking back up as the year closed. Summer vacations across the industry are a likely culprit both times, since fewer people are around to sign new contracts or expand usage.
A separate, broader government survey adds useful context. The Census Bureau's own biweekly business survey puts total AI use across the American economy at roughly 22 percent, far below Ramp's 56 percent. That gap exists because Ramp's customer base skews toward younger, more tech-comfortable companies. It means Ramp's numbers are a useful early signal, but not a mirror of the whole economy.
What looks different this time is what happened at the top of the market. Spending per employee among the heaviest AI users, the top slice of companies that treat AI as central to how they work, fell nearly 10 percent in August. Some of that is the vacation effect. But a lot of it is simple economics: OpenAI and Anthropic have been cutting prices to win business from each other, and the average cost of AI tokens has fallen sharply since its peak earlier this year. When something gets cheaper and usage does not rise fast enough to make up for it, total spending falls even if more people are using the product.
That price war has not yet been rescued by a rush toward cheaper open alternatives. Only about 6 percent of AI-spending businesses on Ramp currently buy access to open-source style models through inference platforms, so that is not the main force behind falling prices. The main force is straightforward competition between the two largest AI companies.
The timing makes this worth more attention than a normal monthly wobble. OpenAI and Anthropic have both filed confidentially to go public this year, with reported valuations approaching or above 800 billion dollars and even a trillion dollars. At the same time, the five biggest cloud and AI infrastructure companies plan to spend somewhere around 660 to 690 billion dollars building data centers and buying chips in 2026 alone. A draft Treasury Department report has already warned that if AI growth slows while spending at that scale continues, the fallout could spread through markets the way the dot-com crash did two decades ago.
None of this means the AI boom is ending. It means the industry's biggest spenders are getting more selective, paying less per person while still paying, and picking older, cheaper models for jobs that do not need the newest and most expensive one. If your company already runs on AI tools, that is good news: prices are falling and there is no reason to lock into an expensive contract right now. If you are watching this as a warning sign for the broader AI investment boom, the real test is whether this slowdown reverses by December like it did last year, or whether it turns into a longer stall.