Businesses are using more AI, but their spending is no longer rising at the same pace. Ramp’s data points to a widening gap between growing token volumes and much slower spending growth, creating a challenge for companies whose business plans depend on rapid revenue expansion.
More AI activity, slower spending growth
Across August and September, AI token spending grew just 1.5%, according to Ramp. In July, it had grown 42% from the previous month. Spending picked up again in October, rising 8.9% from September 23 to October 23, but Ramp’s lead economist, Ara Kharazian, said strong headwinds remain.
The slowdown does not mean companies have stopped using AI. Ramp’s figures show rising volumes, suggesting businesses are processing more tokens even as their bills grow more slowly. That distinction matters: rising usage can support a technology’s adoption without delivering the same rate of revenue growth to the companies selling it.
The biggest AI spenders may be losing momentum
The most AI-intensive firms have been a major source of spending. Kharazian said the top 1% of companies had begun to ease back slightly in August and September, falling just below the $10,000 per-employee level they reached at their summer peaks. He said that change among the largest spenders could be enough to slow revenue growth, even while overall AI volumes rise.
Spending levels vary widely across companies. Median firms spend $12-$15 per employee per month, while the top 10% spend $600-$1,000. Kharazian questioned whether the typical use cases outside the most AI-intensive businesses could grow to match the top 1%’s spending. AI coding agents may justify high outlays in software and finance, but a doctor’s office or car dealership may not need to spend at the same level.
That concentration creates a vulnerability for AI providers. If a relatively small group of customers accounts for much of the spending, even a modest slowdown among them can weigh on growth across the market. Kharazian said almost all AI spending comes from venture-backed, high-growth technology companies, which makes a shared pullback potentially consequential for those businesses.
Competition and efficiency are lowering bills
Ramp’s data points to a buyer’s market, where businesses can seek lower prices and switch among more options. OpenAI and Anthropic have cut prices, while both have also offered enterprise discounts intended to secure customer spending. OpenAI cut API prices in half for GPT-6 Sol and GPT-6 Luna in September. Anthropic reduced pricing for Fable 5.1 and Opus 5.5; the company said in a recent filing that the decision reduced revenue growth by about $600 million.
Customers can also turn to alternatives. Kharazian cited enterprise offerings from Meta and Nvidia, cheaper or open-source models, and model companies in China and the United States. He also pointed to Jev, a lightweight tool that selects between a few predetermined outputs. For tasks that need only a simple yes-or-no answer, routing work to a cheaper option can lower costs further.
Businesses are also learning to manage their AI use. More experienced teams can avoid sending long-running tasks to expensive models, set employee budgets and route work from standard or light models to cheaper ones. Together, lower prices and more deliberate choices mean a company can use more AI while spending less per task.
Investors are watching the growth outlook
The spending trend lands as investors assess the potential valuations and fundraising plans of major AI companies. Bloomberg reported Friday that Anthropic’s potential valuation had fallen from the $2 trillion estimated in early October to $1.2 to $1.3 trillion. The company was also reportedly expected to receive a high “junk” rating from credit agencies. OpenAI lowered its proposed valuation to $1.25 trillion from $1.4 trillion and cut its latest fundraising goals by about $10 billion.
Those developments do not prove that slower spending caused changes in valuations or fundraising. But Ramp’s figures help explain why investors may be questioning whether the industry can maintain the growth pace that once seemed likely. More tokens being used does not automatically mean providers will earn more if customers pay less or find ways to reduce the cost of each workload.
New products could change the picture
One possible source of renewed growth is the arrival of new AI products and use cases. AI coding emerged a year and a half ago as a commercially viable product, and investors are looking for another category that could drive spending higher. Personal agents are one candidate: OpenAI, SpaceXAI and Meta have released their own, alongside startups like Instinct.
Kharazian is skeptical that personal agents will deliver the same economic value as coding tools. He said personal agents are “not as economically productive as coding,” and that the products and commercially advanced uses observable today point to slowing AI spending. For now, the central question is whether future products can create enough new demand to offset lower prices and more efficient use.