AI spending growth among major companies decelerated in August, with the top 1% of AI-using firms reducing their per employee expenditure by nearly 10%, reflecting both seasonal effects and broader market shifts such as declining token prices and slower adoption of frontier models.
- AI adoption growth slowed to 0.4% in August among Ramp clients.
- Top 1% of AI-using firms cut AI spend per employee by nearly 10%.
- Falling token costs and cheaper models impact overall AI expenditure.
What happened
Data from payment processor Ramp illustrates that AI spending growth among 70,000 companies showed a marginal increase of just 0.4% in August, suggesting a slowdown after months of rapid adoption. Notably, the top 1% of AI-using firms decreased their AI spend per employee by approximately 10%, dropping the average to about $7,205. This marks a significant contraction in spending intensity at the tier of companies expected to lead growth in AI usage.
This temporary cooling comes amid seasonal factors such as summer vacations but also coincides with a notable decrease in average token costs for AI services. Prices for tokens dropped from a peak of $1.15 per million in March to $0.68 per million, as leading AI providers like OpenAI and Anthropic lowered their rates. Additionally, many companies are opting for older, more affordable AI models instead of investing in the newest, more costly frontier releases.
Why it matters
The AI infrastructure buildout has been heavily funded on expectations of robust and sustained growth in AI usage, particularly from frontier labs and hyperscalers investing billions in technology and chips. A slowing in adoption and reduced AI spending could jeopardize revenue forecasts crucial for recouping these investments. The decline in token prices, while beneficial for broad accessibility, also indicates that revenue growth is not keeping pace with cost reductions, potentially tightening margins for AI model creators.
Ramp’s data, while not fully representative of the entire market due to its tech-heavy clientele, is one of the few real-time indicators available that can signal shifts in the AI adoption curve. These early warnings suggest that the market dynamics are evolving, with competition driving down costs and shifting usage patterns, especially as companies prioritize user-friendly AI tools over advanced but expensive frontier models.
What to watch next
Stakeholders should monitor AI spending trends beyond the summer lull to distinguish between seasonal effects and more enduring market developments. In particular, it will be critical to see if the demand for frontier AI models rebounds or if companies continue favoring older, cost-effective alternatives, which could limit future revenue growth at AI labs. Industry adoption surveys and token pricing will also provide important signals about the health of AI investments.
Another area to watch is the growing, but still small, use of model-serving or inference platforms, currently used by only 6.4% of AI-spending businesses. Their uptake could influence the broader ecosystem’s growth. Finally, the shift toward accommodating less technical users with coworking AI tools may accelerate, as labs try to broaden their customer base and sustain spending levels despite the challenges posed by price competition and slower enterprise adoption.