Glonce

Ramp data shows AI token spend grew just 1.5% in August and September

businessThe Glonce editorsTechCrunch🔥59Glonce rating59Impact55Novelty40Relevance80Surprise50Players70Credibility75
Ramp data shows AI token spend grew just 1.5% in August and September

A new report from Ramp shows that companies are using more AI than ever but are not spending more on it. According to Ramp's data, token spend grew only 1.5% across August and September combined. In July it had grown 42% from the previous month. TechCrunch reads this as a sign that price competition and savvier customers may be capping revenue for OpenAI and Anthropic, and says the data suggests investors might be wrong to assume AI companies will keep growing at breakneck speed as Anthropic gears up for its IPO.

Ramp's lead economist, Ara Kharazian, told TechCrunch that almost the entire spend on AI comes from venture-backed, high-growth tech companies. That is fine while spend keeps rising, he said, but if it slows because of something in the market, they will all slow at once, which could lead to cascading effects. He sees the top 1% of firms, the most "AI-pilled", stuttering slightly in August and September, coming in just below the $10,000 per employee they were hitting at their summer peaks. A slowdown in that cohort is, in his words, "sufficient to slow down revenue entirely, despite the fact that we're seeing rising volumes." Ramp data shows spending growth among median firms and the top 10%, but not enough to make up the difference. Median firms spend $12-$15 per employee per month, and the top 10% spend $600-$1,000.

Kharazian doubts that today's use cases will push most firms toward the $10,000 per employee per month mark the top 1% is at. Demand for AI coding agents is strong in software and finance, he said, but it is not necessarily true that a doctor's office or a car dealership will need to spend that much. "The era of hypergrowth may be over," he said, adding that there may not be much more to squeeze out of the top 1%. October data did show a re-acceleration, with token spend up 8.9% from September 23 to October 23, but he said the headwinds on spend growth are strong.

He says the slowdown is not a demand problem, since volumes are up. He points instead to a competitive market. In September, OpenAI cut API prices in half for GPT-6 Sol and GPT-6 Luna. Anthropic answered with significant price reductions for Fable 5.1 and Opus 5.5, a decision the company said in a recent filing slashed revenue growth by about $600 million. Both companies also offer enterprise discounts meant to lock in spend, but TechCrunch writes that they are in reality hitting their revenue targets. Kharazian calls it very much a buyer's market. TechCrunch notes that subsidising a pricey technology to lock in customers is an old Big Tech playbook, as with early Uber and Lyft fares, but Kharazian says that does not work in AI because there is too much competition.

He lists the competition: Meta and Nvidia are introducing enterprise offerings with pretty cheap models or open source models; there are other open model companies from China and the United States; and a new company called Jev, a lightweight tool for choosing between a few predetermined outputs, is increasingly among the top performers on Ramp. "For a lot of AI use cases, you just need a simple Yes/No," he said, so cheap models get shifted to Jev and get cheaper still.

The second factor is thrift. Kharazian says advanced AI users get better at controlling spend: they set defaults that block very expensive models for long-running tasks, impose budgets on employees, and use routers to offload work from standard and light models to even cheaper light ones. Notion co-founder Akshay Kothari recently posted internal data that matched Ramp's insights, putting it as "tokens go brrr. costs go ahhh."

The market backdrop is gloomy. Bloomberg reported Friday that Anthropic's potential valuation sank from the $2 trillion estimated in early October to $1.2 to $1.3 trillion, and that the company is 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. Kharazian says Ramp's dataset likely informs many investor concerns.

There is a speculative bright side. The hypergrowth thesis rests on AI introducing new use cases. AI coding emerged only a year and a half ago as the most commercially viable product and led Anthropic and OpenAI to become trillion-dollar companies, and investors hope for another revolution. Some think personal agents will be the next coding agents; OpenAI, SpaceXAI and Meta have released their own, alongside startups like Instinct. Kharazian is skeptical, since personal agents are "not as economically productive as coding." Based on what can be observed today, he said, AI spend has started to slow down.

Key facts

  • Ramp data: token spend grew only 1.5% across August and September combined, against 42% month-on-month growth in July, while token volumes kept rising.
  • Ramp's lead economist Ara Kharazian says the top 1% of firms stalled just below $10,000 per employee, enough to stall overall revenue growth; median firms spend $12-$15 per employee per month and the top 10% spend $600-$1,000.
  • He blames price cuts (OpenAI halved API prices for GPT-6 Sol and GPT-6 Luna in September; Anthropic cut prices for Fable 5.1 and Opus 5.5, which it said in a filing cost about $600 million of revenue growth), cheap rival models and thriftier customers.
  • October showed a re-acceleration, with token spend up 8.9% from September 23 to October 23, but Kharazian says headwinds remain strong.
  • Bloomberg reported Anthropic's potential valuation fell from $2 trillion in early October to $1.2 to $1.3 trillion; OpenAI cut its proposed valuation to $1.25 trillion from $1.4 trillion.

Why it matters

Investors are pricing Anthropic's IPO and OpenAI's fundraising on the assumption that AI revenue keeps growing at breakneck speed. Ramp's numbers show spend growth falling from 42% in July to 1.5% across August and September, while usage kept climbing. Kharazian's reading is that people can use more AI but use it for less, which is good for customers and raises new questions about the model companies' revenue trajectories. The worry is also about concentration: almost all AI spend comes from venture-backed, high-growth tech firms, so a slowdown could hit them all at once.

Who it affects

Model companies first: OpenAI and Anthropic, whose price cuts (in Anthropic's case, about $600 million of revenue growth by its own filing) sit alongside enterprise discounts that aim to lock in spend while the companies are still hitting their revenue targets. Investors weighing the Anthropic IPO and OpenAI's raise are next. Customers benefit from a buyer's market and cheaper tokens. The highest-spending firms, the top 1%, drive the effect most, and cheaper rivals such as Meta and Nvidia offerings, open models and Jev gain from the shift.

How to use it

For companies buying AI, Kharazian describes what advanced users do to control cost: set defaults that prevent very expensive models from running long tasks, impose budgets on employees, and use routers to push work from standard and light models to even cheaper light models. For simple Yes/No style tasks he points to lightweight tools such as Jev. For anyone tracking the sector, the figures to watch are spend growth against volume growth, and whether the October re-acceleration of 8.9% holds.

How solid is it

The core numbers (1.5%, 42%, 8.9%) are Ramp data as relayed by TechCrunch; the article does not give the Ramp report's title, publication date, sample size or number of customer firms, and it gives growth rates rather than absolute token volumes or dollar totals. The causes (price cuts, discounts, thrift) come from Kharazian's interpretation and TechCrunch's own framing. The roughly $600 million figure is Anthropic's own statement in a recent filing, though the article does not say which filing or period. The valuation and credit-rating details are Bloomberg's, secondhand. Notion co-founder Akshay Kothari's internal data is cited as matching Ramp's findings.

Risks and caveats

One month of re-acceleration (8.9% from September 23 to October 23) complicates the flatline story. Kharazian also calls the next-revolution hope speculative: new products such as personal agents, which OpenAI, SpaceXAI and Meta have released alongside startups like Instinct, could restart growth, though he is skeptical they will match coding. The Anthropic junk rating is reported as expected, not confirmed. No IPO date is given; the article says only that Anthropic is gearing up for one. Kharazian has published a series of blogs called "Cracks in the AI Thesis", so his view on the topic is long-standing.

“You can use more AI, but use it for less.”

— Ara Kharazian, Ramp's lead economist