Business AI spending cooled: companies use more AI than ever and pay almost the same
Direct answer
According to Ramp data reported by TechCrunch on November 6, 2026, business spending on AI tokens grew 42% in July and only 1.5% across August and September combined, even as token volume kept rising. October brought a partial recovery, with spending up 8.9% between September 23 and October 23, while volume rose 37.1%. Ramp's lead economist Ara Kharazian reads it plainly: demand kept rising; what changed was the price. Models got cheaper and companies learned to spend more carefully. For AI buyers, that is good news: the negotiating power is on your side.
The numbers, in one place
- July 2026: up 42% month on month. The last month of hypergrowth in token spending, according to Ramp's data.
- August and September combined: up 1.5%. Two months of spending essentially flat, even as token usage kept growing.
- September 23 to October 23: spending up 8.9%, volume up 37.1%. Consumption is growing much faster than what is paid for it.
- The blended price per million tokens fell from $0.76 to $0.56 between those two weeks, a 26.8% drop: OpenAI down 42.4% and Anthropic down 15.7%, according to Ramp's index.
- The median firm spends $12 to $15 per employee per month. The top 10% pay $600 to $1,000 per employee. The most intense 1% reached close to $10,000 per employee in the summer and dipped slightly in August and September.
- Spending is concentrated: the top 1% of firms drive 80% of spending on OpenAI and Anthropic. And almost all of the spend comes from venture-backed technology companies, not from the economy at large.
Why AI got cheaper (and what to do about it)
- A price war between the labs: in September, OpenAI halved API prices for two of its GPT-6 models, and Anthropic cut its own prices, including Fable 5.1 and Opus 5.5. Anthropic said in a filing that the decision cut its revenue growth by about $600 million.
- Customers learned to route: simple tasks move to cheap models and expensive models are reserved for work that truly needs them. Usage rises, spend per task falls.
- Anthropic gained share while the pie cooled: its share of token spending rose from 51.3% to 60.5% between September 25 and October 23, while OpenAI fell from 44.2% to 35.1%. The context is its IPO: Bloomberg reported its indicated valuation fell from about $2 trillion in early October to $1.2–$1.3 trillion.
- What to do if you buy AI: renegotiate with the data in hand, ask for pricing per task rather than per token, and separate the experimentation budget from the production budget. A vendor charging you for an expensive model on a task a cheap model solves just as well is overcharging you.
Note: this article summarizes Ramp data reported by TechCrunch on November 6, 2026 ("Ramp data shows AI spending stalling just as Anthropic nears its IPO", by Rebecca Bellan) and Ramp's index, with coverage from Unite.AI and other outlets. Ramp's data comes from more than 85,000 U.S. businesses through its corporate-card platform; token metrics cover a narrower subset and do not represent all of Ramp's AI spend. Anthropic's valuation figures are reported by Bloomberg, not confirmed by the company.
Want AI that works for your business?
Message us on WhatsApp and let's talk about automation and AI agents for your business.
WhatsApp inicIARelated articles
- ROI of artificial intelligence: the real numbers for 2026
- How much implementing AI costs a business in 2026 (and the 3 mistakes that burn the most money)
- Small-business AI spending soared and Anthropic took the crown from OpenAI: your advantage as a buyer
Frequently asked questions
Is business AI spending falling?
It is not falling: it cooled. It grew 42% in July 2026 and only 1.5% across August and September, with a partial 8.9% recovery in October. Token usage kept rising the whole time, up 37.1% in the latest measured period.
Why spend less if companies use more AI?
Because prices fell: the average price per million tokens dropped from $0.76 to $0.56 in a month. OpenAI and Anthropic cut prices, and companies learned to route each task to the cheapest model that solves it.
How much does a typical company spend on AI?
According to Ramp, the median spends $12 to $15 per employee per month on tokens. The top 10% pay $600 to $1,000 per employee, and the most intense 1% reached close to $10,000 per employee in the summer of 2026.
What does this mean for AI buyers?
Negotiating power is on the buyer's side: renegotiate with the data in hand, ask for pricing per task rather than per token, and separate the experimentation budget from the production budget.
