Ramp’s AI Index, which uses transaction data from more than 85,000 American businesses on the corporate card and spend management platform, shows total AI token volume up 37.1% in the week ending Oct. 23, 2026, compared with the same week a month earlier. Token spending rose 8.9% over the same period.
The spending increase follows a late-summer stretch in which Ramp said token spend recorded two straight months of virtually no growth across August and September even as token volumes rose.
In September 2026, 55.9% of U.S. businesses on Ramp had observed spending on AI models, subscriptions, tokens, or APIs, up 0.6 percentage points from August and 12.1 points from a year earlier. In Ramp’s token dataset, Anthropic held the largest share of token spending in the week ending Oct. 23 at 60.5%, while OpenAI had 35.1%.
Price competition is changing the revenue math
The blended price per million tokens across OpenAI and Anthropic fell 26.8% to US$0.56 in Ramp’s token data, down from US$0.76 a month earlier. OpenAI’s blended price fell 42.4%, while Anthropic’s fell 15.7%.
The same pricing pressure appears in company announcements and filings. OpenAI said in a September update that API prices for GPT-6 Sol and GPT-6 Luna were 50% lower than GPT-5.6 promotional pricing. Anthropic said in an amended SEC filing that Fable 5.1 reduced typical workload costs by about 25% compared with Fable 5, and Opus 5.5 reduced customers’ cost per task by about 40% compared with Opus 5.
Anthropic also said its revenue increase from the second quarter to the third quarter of 2026 was partially offset by approximately US$600 million from price reductions introduced with new model versions and contractual discounts.
The top-spending customers carry the risk
Ramp’s September data shows a huge spending divide between ordinary AI buyers and the heaviest users. Among firms paying for AI, the top 1% spent a median of US$9,567.53 per employee per month on AI. The median firm spent US$15.70.
Ramp’s October analysis said the top 1% cohort was still below its July high, even after rebounding from August. That matters for model providers because heavy enterprise users account for a large share of AI revenue.
TechCrunch, which interviewed Ramp lead economist Ara Kharazian, reported that AI spending is heavily concentrated among venture-backed, high-growth tech companies. Kharazian told TechCrunch that growth from median firms and the top 10% was not enough to offset a slowdown among the top 1%.
The finding complicates the revenue case for model providers that depend on expanding enterprise token use, especially in coding and agentic workflows. It also suggests that wider AI adoption does not automatically translate into proportional revenue growth for the companies selling the models.
Buyers may be getting better terms
In Ramp’s analysis, slower spending growth reflects falling prices, model competition, enterprise discounts, and more disciplined controls over which models employees use for which tasks.
That can be positive for business users. Lower token prices and cheaper standard models can make AI workflows less expensive to test, especially where frontier models are unnecessary. The tradeoff is that spend alone becomes a weaker proxy for adoption or value. A company can use more AI while paying less, but that says little about whether the output is saving time, increasing revenue, reducing errors, or improving customer experience.
For marketing, content, and SEO teams, this reinforces a practical point Tech Help Canada has made in its guidance on using AI tools for SEO without publishing low-value content: AI should be measured by the quality and usefulness of the work it supports, not by how much software spend appears on the card statement.
The data has limits
Ramp’s AI adoption figures are based on observed payments by Ramp customers, not a census of all U.S. firms. Its tokenomics data comes from businesses using Ramp’s Token Spend Management product, and Ramp says that sample is not representative of the company’s total AI spend volume, although results are directionally similar.
The index also cannot separate every cause behind lower spending growth. Seasonality, discounts, model routing, cheaper default models, and tighter employee budgets can all reduce spending without reducing actual AI usage. A slowdown in spending growth is therefore not the same as a collapse in AI use.
I can suggest related Tech Help Canada articles based on the topic you’re reading now.
Tech Help Canada Staff researches, writes, and reviews practical content for business owners and professionals. Our coverage spans business, marketing, SEO, technology, and the tools and systems people use to grow and operate online. We focus on clear, useful information backed by research, hands-on experience, and editorial review. Learn more about our team and editorial standards. Need help with something? Contact Us
