Top AI spenders cut per-employee spend 9.7% in August as token prices fall 41% from March peak

Ramp data covering 70,000 companies shows August AI adoption rose just 0.4% month-over-month (56% of customers paid for AI products) while per-employee spend at the top 1% of spenders fell 9.7% to $7,205, as effective token prices dropped 41% from the March…

This is the first merged summary for this story (no prior baseline). Ramp's September 2026 AI Index, covering 70,000 companies, found that 56% of its customers paid for AI products in August, up only 0.4% month-over-month (TechCrunch). Per-employee AI spending at the top 1% of spenders fell 9.7% in August to $7,205 (The Decoder; TechCrunch describes it as a drop of nearly 10%). The Decoder attributes the decline partly to August vacations, falling token prices, and migration to cheaper models. Effective token prices fell 41% from their March 2026 peak to $0.68 per million tokens (The Decoder); TechCrunch pegs the peak at $1.15 and attributes the cuts to price reductions by OpenAI and Anthropic, noting the labs have not yet offset the cuts with volume. Frontier models (Opus, Fable, Sol) fell from 53% of tokens consumed in early August to 45% by early September (The Decoder), while TechCrunch reports customers shifting to cheaper models such as ChatGPT 5.6-Terra and Claude Sonnet. On vendor adoption, Anthropic was paid for by 43.8% of US companies (up 0.34 points) versus 39.8% for OpenAI (up 0.09 points), with IT and finance leading adoption (The Decoder). Broader context from TechCrunch: the US Census Bureau's survey shows only 22% of US businesses report using AI, suggesting Ramp's tech-heavy client base overstates adoption. Note a discrepancy: both outlets cite a 6.4% figure but attach it to different metrics — TechCrunch says only 6.4% of AI-spending businesses used model-serving or inference platforms in August, while The Decoder says only 6.4% of AI-using companies on Ramp run open-weight models.

• Ramp's September 2026 AI Index covers 70,000 companies (TechCrunch).
• 56% of Ramp customers paid for AI products in August, up 0.4% month-over-month (TechCrunch).
• Per-employee AI spend at the top 1% of spenders fell 9.7% in August to $7,205 (The Decoder); TechCrunch characterizes the same figure as a drop of nearly 10%.
• The Decoder attributes the spending decline partly to August vacations, falling token prices, and migration to cheaper models.
• Effective token price dropped 41% from its March 2026 peak to $0.68 per million tokens (The Decoder); TechCrunch reports the peak as $1.15 per million and attributes the price cuts to OpenAI and Anthropic.
• TechCrunch reports the labs have not yet offset the price cuts with volume.
• Frontier models (Opus, Fable, Sol) fell from 53% of tokens consumed in early August to 45% by early September (The Decoder).
• TechCrunch reports customers shifting to cheaper models such as ChatGPT 5.6-Terra and Claude Sonnet.