Business investment in AI tools showed signs of slowing in August, according to spending data from Ramp, which tracks payments across 70,000 companies. The latest figures reveal that 56% of Ramp’s customers paid for AI products in August, a marginal increase of 0.4% from July. This follows a similar plateau observed last year between August and October before adoption accelerated again later in the year.
The rapid expansion of AI infrastructure by leading labs and hyperscalers depends on continued growth in usage and revenue. While adoption has surged, particularly among software engineers using agentic coding tools, any slowdown could impact the financial returns on these substantial investments.
Ramp’s data may overrepresent AI adoption due to its tech-focused customer base. For comparison, a recent US Census Bureau survey found only 22% of businesses reporting AI use as of late August. Despite this, Ramp’s spending data remains one of the few direct indicators of AI adoption trends.
August’s slowdown may partly reflect seasonal factors, as many industry professionals take vacations during this period. However, Ramp economist Ara Kharazian points to a significant 10% drop in AI spend per employee among the top 1% of firms, falling to $7,205. This decline also coincides with reduced token costs, as providers like OpenAI and Anthropic have lowered prices from a peak of $1.15 per million tokens in March to $0.68.
The data suggests that price reductions have not yet been offset by increased usage volumes. Many customers appear to prefer older, less expensive AI models such as OpenAI’s ChatGPT 5.6-Terra and Anthropic’s Sonnet over newer, more advanced releases. Since much of the cost of training new models is recouped shortly after release, slower adoption could affect the economics for AI developers.
Additionally, only 6.4% of AI-spending companies used model-serving or inference platforms in August. Although this segment is growing steadily, it is not expanding quickly enough to drive broader adoption across businesses.
Kharazian notes that competition between OpenAI and Anthropic is making AI tools more accessible and affordable, which is reducing spending among the highest-spending firms previously expected to lead growth. This trend may explain why AI labs are increasingly targeting non-technical users with collaborative AI tools.
While this data point could be viewed as a temporary dip or a warning sign for AI model developers and hyperscalers with large hardware investments, it remains positive for companies actively using AI. The situation underscores the evolving dynamics in AI adoption and the challenges in sustaining rapid growth amid changing market conditions.