Companies are using more AI than ever, but they are not paying more for it. New data from Ramp, a company that tracks business spending, shows AI spending grew just 1.5% across August and September. In July it grew 42% over the month before. Usage kept rising during that stretch.
What changed
AI got cheaper. In September, OpenAI cut some of its API prices in half, including for GPT-6 Sol and GPT-6 Luna. An API is how a company plugs AI into its own software. Anthropic also cut prices for Fable 5.1 and Opus 5.5. In a recent filing, Anthropic said that cut reduced its revenue growth by about $600 million.
Ramp's lead economist, Ara Kharazian, says it is now a buyer's market. Meta and Nvidia are offering cheaper models, and open models from Chinese and US companies add more competition. A tool called Jev, built for simple choices like yes or no, is also climbing in Ramp's data.
Customers are getting smarter
Price cuts are only part of it. Kharazian says companies that get good at AI also get better at controlling the bill. Common moves include:
- Defaults that block very expensive models on long-running tasks
- Budgets for individual employees
- Routers that send simple jobs to cheaper models
Notion co-founder Akshay Kothari posted internal numbers that matched Ramp's findings. His summary: "tokens go brrr. costs go ahhh." Tokens are the small chunks of text AI models count and charge for.
Why the top spenders matter
Kharazian says almost all AI spending comes from venture-backed, high-growth tech companies. The top 1% of firms were spending close to $10,000 per employee at their summer peaks. They dipped slightly in August and September. Median firms spend $12 to $15 per employee per month, and the top 10% spend $600 to $1,000. He doubts a doctor's office or car dealership will ever spend like the top 1%.
What it means for investors
The timing is awkward for Anthropic, which is preparing an IPO. Bloomberg reported that its possible valuation fell from $2 trillion in early October to $1.2 to $1.3 trillion. It also reported that credit agencies are expected to give the company a high-risk "junk" rating. OpenAI lowered its proposed valuation to $1.25 trillion from $1.4 trillion and cut its fundraising goal by about $10 billion.
What happens next
The picture is not settled. Spending rebounded in October, up 8.9% from September 23 to October 23, though Kharazian says the headwinds are strong. For business owners, the lesson is practical. Prices are falling, so set budgets and keep pricey models for the hard jobs.
The takeawayAI is getting cheaper, so set budgets and use costly models only for hard tasks, while simple jobs go to cheaper ones.