The breakneck pace of corporate AI adoption hit a speed bump in August. Spending data from Ramp, covering 70,000 companies, reveals that only 56% of its customers paid for AI products last month. That is a meager 0.4% increase from July. While the absolute number remains high, the momentum has clearly stalled. This isn’t unprecedented; Ramp’s index showed similar stagnation between August and October last year before rebounding. But given the massive infrastructure bets made by hyperscalers and frontier labs, even a slight deceleration raises eyebrows. The entire AI buildout relies on the assumption that usage will keep climbing to justify the billions spent on chips and data centers.
Digging deeper, the story gets more nuanced. Ara Kharazian, an economist at Ramp, points to a nearly 10% drop in AI spend per employee among the top 1% of AI-using firms. Their average spend fell to $7,205. Part of this is likely seasonal, with many tech workers on vacation in August. But it also reflects a structural shift: token prices are plummeting. Average costs have dropped to $0.68 per million tokens, down from a peak of $1.15 in March. OpenAI and Anthropic are engaged in a price war, making AI cheaper but reducing total revenue volume. Companies aren’t necessarily using less AI; they are just paying less for it. Many are opting for older, cheaper models like Anthropic’s Sonnet instead of the latest flagship releases.
This dynamic threatens the economic model of frontier labs. Industry insiders suggest that much of the training cost for new models is recouped in the first few weeks after launch. If adoption slows or users stick to cheaper legacy models, that quick payback window shrinks. Meanwhile, the rise of open-weight models hasn’t yet tipped the scales. Only 6.4% of AI-spending businesses used model-serving platforms in August. That share is growing, but it is not yet a major factor in broader business adoption. The competition between OpenAI and Anthropic is driving accessibility, but it is also driving down the very revenue growth the market expected from heavy users.
For now, this might just be a summer lull. But if you are a chipmaker or a cloud provider with hundreds of billions in hardware orders, a plateau in spending is a warning sign. For everyday businesses, however, the news is positive. AI is becoming more affordable and accessible. The question remains whether lower prices can stimulate enough new volume to offset the margin compression. If not, the AI boom may need to find a new gear to sustain its current valuation multiples.
