Ramp data shows corporate AI spending flattening as token volumes keep rising

Flat spending despite rising usage

Companies are consuming more AI than ever, but the money flowing to AI vendors has stopped growing alongside it. According to TechCrunch, a new report from spend-management firm Ramp shows token spending grew only 1.5% across August and September combined, a sharp deceleration from July, when it rose 42% month over month. Usage volumes kept climbing through the same stretch, which points to falling prices rather than fading demand.

October brought a partial rebound, with Ramp measuring an 8.9% rise in token spend between September 23 and October 23. Still, Ramp's lead economist Ara Kharazian told TechCrunch that the pressures holding back spending growth remain powerful.

A narrow base of big spenders

Ramp attributes much of the stall to concentration. Kharazian said nearly all AI spend comes from venture-backed, high-growth technology companies, and that a market-wide shock could slow the entire cohort at once, with cascading effects. The heaviest users, the top 1% of firms by AI spend, slipped just below the $10,000-per-employee mark they hit at their summer peak, and weakness in that group alone was enough to drag overall revenue growth down despite rising volumes.

Growth continued among median firms, which spend $12 to $15 per employee per month, and the top 10%, which spend $600 to $1,000, but not enough to offset the leaders. Kharazian also questioned whether typical businesses will ever reach top-cohort spending levels, noting that while software and finance have strong appetite for coding agents, a doctor's office or car dealership may simply not need that much AI.

A buyer's market in models

The slowdown is not a demand problem, Kharazian argued, but a pricing one. In September, OpenAI halved API prices for GPT-6 Sol and GPT-6 Luna, and Anthropic answered with significant cuts to Fable 5.1 and Opus 5.5, a move the company disclosed in a recent filing reduced its revenue growth by roughly $600 million. Both vendors have also been offering enterprise discounts intended to lock in spend, though these are denting revenue targets instead.

Kharazian told TechCrunch that AI has become a buyer's market, and that the classic Big Tech playbook of subsidising usage to corner a market does not work here because competition is too broad. Meta and Nvidia are pushing enterprise offerings built on cheap or open models, open-model companies in the US and China are multiplying, and a startup called Jev, a lightweight tool that picks between a few predetermined outputs, is climbing Ramp's rankings because many workloads only need a simple yes-or-no answer routed to something inexpensive.

Customers are also getting thriftier as they mature, Kharazian said, using defaults that avoid expensive models on long-running tasks, capping employee budgets and routing workloads from standard models down to cheaper light models. Notion co-founder Akshay Kothari published internal data that matched Ramp's picture, summing it up as token usage exploding while costs shrink.

Valuations feel the squeeze

The timing is awkward for the model makers. Bloomberg reported that Anthropic's potential valuation dropped from roughly $2 trillion in early October to between $1.2 and $1.3 trillion, and that the company is expected to receive a high junk rating from credit agencies. OpenAI, meanwhile, cut its proposed valuation from $1.4 trillion to $1.25 trillion and trimmed its latest fundraising goals by about $10 billion.

As for the next leg of growth, Kharazian is skeptical that personal agents, released by OpenAI, SpaceXAI and Meta alongside startups like Instinct, will match coding's commercial impact, since they are not as economically productive.

Why it matters

Rising token volumes paired with flat revenue is a classic deflation signal, and it lands at the worst possible moment for vendors: Anthropic is heading toward an IPO while its implied valuation slides. If the bulk of AI spend really does sit with a small pool of venture-backed tech firms, the sector's growth story depends on a narrow, fragile base, and investors underwriting trillion-dollar valuations may be pricing in a trajectory the spending data no longer supports. The open question is whether a new killer use case arrives before the price war finishes eroding the current one.