Ramp AI Index, September 2026: Cracks in the AI Thesis, Part 2
Ramp's AI Index for September 2026 (September 2026; transactions from 70,000+ U.S. businesses on its platform, usage data) found that 43.8% of businesses paid for Anthropic and 39.8% for OpenAI in August, while median AI spend per employee at the top 1% of spenders fell 9.7% to $7,205 a month.
Key findings
- 0143.8% of U.S. businesses on Ramp paid for Anthropic subscriptions or tokens in August, up 0.34 percentage points.OpenAI rose only 0.09 points to 39.8%. Ramp says overall new AI adoption is still growing but decelerating.
- 02Median per-employee monthly AI spend among the top 1% of spenders fell 9.7%, from $7,976 to $7,205.Ramp notes the top-1% figure is volatile, was revised up for July, and may partly reflect August vacations.
- 03Ramp's effective price per million tokens fell 41% to $0.68, from a 2026 peak of $1.15 in March.Cheaper standard models are driving volume growth, which may not offset lower prices.
- 04Frontier models drove 45% of token share, down from a 53% peak in August.Ramp reports hearing from businesses that set company-wide defaults steering staff to cheaper standard models.
- 05Only 3.6% of all businesses (6.4% of AI-spending businesses) use open-source models.Ramp measures this through routing platforms that also offer closed models, so it says actual open-source use is likely lower.
By the numbers
43.8%of businesses paid Anthropic in August 2026
39.8%of businesses paid OpenAI in August 2026
-41%change in effective price per million tokens since March 2026 peak
What it means for you Draft
For executives at $10–100M companies
This is spending data from real card and bill transactions, not a survey, so it shows what businesses actually pay for. Falling token prices and company defaults to cheaper models suggest AI costs per employee can be managed down without giving up the tools. A $10–100M company may want to check which models its teams default to, since frontier models cost more and may not be needed for routine work.
For practitioners
Set sensible default models for routine tasks and reserve frontier models for work that needs them. Track AI spend per employee monthly; Ramp's own figures show it can swing sharply month to month.
Limitations
Medium trust.Large transaction dataset with public methodology, but a non-random customer base; Ramp sells spend-management software with AI features.
Ramp's customers skew toward startups and tech-forward firms, so adoption rates likely run higher than for U.S. businesses overall. The top-1% spend figure is volatile and subject to revision, and the letter does not publish sample counts for the month.
About the publisher: Sells corporate cards and spend-management software, including AI features; the index is built from its own customers' transactions.
Ramp. "Ramp AI Index, September 2026: Cracks in the AI Thesis, Part 2." September 9, 2026. https://ramp.com/data/ai-index-sept-2026
Related reports
A New Look at AI's Impact on Jobs: Firm-Level AI Spending and Workforce AdjustmentLinking real AI spending to headcount, Ramp and Revelio Labs find heavy AI spenders grew staff, including entry-level roles, after a 6–12 month lag.
Gartner Survey Finds Only 22% of Organizations Have Successfully Scaled AI Across Multiple Business Units (press release)85% of functional leaders plan to increase AI spending in 2026 after allocating an average 12% of functional budgets to AI in 2025.
Businesses Are Using AI to Transform Work, Not Cut JobsAI use among regional firms jumped again in 2026, but layoffs stayed rare; retraining remains the main workforce response.