AI usage keeps climbing, but the money has stopped following at the same pace. TechCrunch AI reports that token spend in Ramp’s data grew just 1.5% across August and September, after jumping 42% in July alone. That’s an awkward number to land right as Anthropic gears up for its IPO.
October did bounce back, with spend up 8.9% from September 23 to October 23. But Ramp lead economist Ara Kharazian says the headwinds remain strong. He doesn’t see a demand problem, because volumes are still rising. He sees customers getting more AI for less money.
Why more usage isn’t bringing more revenue
Kharazian points to two main forces:
- A price war. OpenAI cut API prices in half for GPT-6 Sol and GPT-6 Luna in September. Anthropic answered with big cuts to Fable 5.1 and Opus 5.5, and said in a recent filing that the decision slashed revenue growth by about $600 million.
- Thriftier customers. Advanced users set defaults that keep expensive models off long-running tasks. They give employees budgets and use routers to move work to cheaper models.
The old Big Tech playbook says you subsidize now and raise prices once rivals are gone. Uber and Lyft rides got a lot more expensive that way. Kharazian doesn’t think it works here, because there’s too much competition. Meta and Nvidia are selling enterprise offerings with cheap or open models, and open model companies in China and the US keep adding pressure.
Then there’s Jev, a lightweight tool for choosing between a few predetermined outputs. “For a lot of AI use cases, you just need a simple Yes/No,” Kharazian said. Already cheap work gets even cheaper.
The 1% problem
The more worrying number is how concentrated this spending is. Here’s Ramp’s monthly spend per employee:
- Median firms: $12 to $15
- Top 10%: $600 to $1,000
- Top 1%: just under the $10,000 peak they hit this summer
That top tier spends roughly 700 times what a median firm does, so a small stumble there can erase growth everywhere else. It already has: spending rose among median and top 10% firms, but not enough to make up the difference.
“Almost the entire spend on AI is from venture-backed, high-growth tech companies,” Kharazian said. If that group slows, he warned, “they will all start to slow down simultaneously,” which could trigger other cascading effects.
The obvious hope is that everyone else catches up. Kharazian isn’t convinced, saying “it’s not necessarily true that a doctor’s office or a car dealership will need to spend that much on AI.”
Investors are already repricing
Bloomberg reported Friday that Anthropic’s potential valuation fell from an estimated $2 trillion in early October to $1.2 to $1.3 trillion. Anthropic is also reportedly expected to get a high “junk” rating from credit agencies. OpenAI lowered its proposed valuation from $1.4 trillion to $1.25 trillion and cut its latest fundraising goals by about $10 billion.
I’d keep some caveats in mind, though. Ramp only sees its own customers. Kharazian also has a clear point of view, having written a blog series called “Cracks in the AI Thesis.” And two soft months followed by a rebound don’t prove a trend.
The bull case is a new category. AI coding emerged only about a year and a half ago and helped turn both labs into trillion-dollar companies. Some think personal agents could be the next wave, but that’s still speculative.
Where this goes over the next 1 to 3 years
These are my projections, not Ramp’s:
- Next 12 months: Public market scrutiny makes spend concentration and pricing power as important as headline growth. Anthropic’s IPO will be the first big test.
- 1 to 2 years: Routing becomes standard, and premium models have to justify their price task by task.
- 2 to 3 years: Growth depends on a new breakout product or on mainstream businesses finding uses worth far more than $15 per employee.
What to do now
- If you buy AI: Use the leverage while it’s a buyer’s market. Be careful with long commitments, since prices have been falling fast.
- If you build with AI: Audit which tasks really need a top model. Simple classification work is the obvious place to start.
- If you sell to AI-heavy startups: Check your exposure to that top 1%, because Kharazian’s cascade would hit you too.
- If you invest: Watch spend per employee at the top firms, along with token volume.
As Notion co-founder Akshay Kothari put it, “tokens go brrr. costs go ahhh.” Great for customers. The open question is how long model companies can grow on those terms, and Anthropic’s IPO should give us the first public answer. The full report is at TechCrunch AI.