- According to Ramp, in the week to October 23, 2026 companies used 37.1% more tokens than a month earlier but paid only 8.9% more for them.
- The blended price per million tokens across OpenAI and Anthropic fell from $0.76 to $0.56, down 26.8%.
In short: Ramp, which runs corporate cards and spend management for thousands of American companies, publishes the Ramp AI Index, a measure of what businesses actually pay for AI. The latest data shows a gap: companies are using much more AI, but their bills are growing far more slowly. TechCrunch covered the numbers on November 6, 2026.
What the data shows
- In the week ending October 23, 2026, companies used 37.1% more tokens than in the week ending September 25, 2026, but spent only 8.9% more on them.
- The blended price per million tokens across OpenAI and Anthropic fell from $0.76 to $0.56, a drop of 26.8%. OpenAI’s fell 42.4% and Anthropic’s 15.7%.
- In August and September token spending barely grew even though usage did. In early October spending reached a new peak, 1.5% above the August high.
- Over the same period Anthropic’s share of token spending among Ramp customers rose from 51.3% to 60.5%.
Why the gap
Ramp’s lead economist Ara Kharazian points to price cuts by the biggest providers and to better, cheaper models across the board. He also writes that enterprise contracts rarely commit customers to higher spending and more often come with discounts off the list price. Speaking to TechCrunch, he called it a buyer’s market and noted that almost all AI spending comes from fast-growing, venture-backed tech companies.
How to read it
This is data from one payments platform and its customers, not the whole market, and Ramp itself says the figures for the heaviest spenders are more volatile. My practical takeaway: every few months, recalculate what your AI tasks cost at current prices, and ask for a discount if you spend a lot. I write about model prices in the Model pricing hub.