Ramp's Data Suggests AI's Top Spenders May Have Learned to Pay Less for More

The AI boom's favorite story — spend more, get more — just hit its first real snag. Ramp's data shows token spending grew just 1.5% across August and September, even as usage kept climbing. The biggest spenders in AI didn't stop using AI. They got better at not paying for it. And the labs, in a fit of competitive generosity, helped them do it.

The top 1% learned to shop

This is where the story lives. One percent of companies account for 80% of business spending on OpenAI and Anthropic. These are the most "AI-pilled" firms in existence, and in September they spent $9,568 per employee per month — still 3.1% below July's peak. The median firm, meanwhile, spends $12 to $15 per employee per month. The entire AI economy rests on a sliver of companies spending a thousand times the median — and that sliver just discovered budgets, model routers, and the radical idea of not running the most expensive model for a yes/no question.

Ramp's lead economist Ara Kharazian put it bluntly: "The era of hypergrowth may be over, and now that we know there may not be as much juice to squeeze out of the top 1%."

The sellers cut their own prices — then watched the cuts eat their revenue

The customers didn't do this alone. OpenAI halved API prices in September for GPT-6 Sol and GPT-6 Luna. Anthropic cut pricing for Fable 5.1 and Opus 5.5 — a decision its own filing says slashed revenue growth by about $600 million. The blended price per million tokens across both labs fell 26.8% to $0.56 in the four weeks to October 23, while token volume jumped 37.1% and spending rose just 8.9%.

Call it price competition or call it a land-grab — either way, the labs are racing to make their own product cheaper, faster than usage can grow. That's a great deal for buyers and a genuinely uncomfortable question for anyone pricing these companies on ever-rising spend.

Anthropic is winning the share war while the pie flattens

In those same four weeks, Anthropic's share of the token spending Ramp tracks rose from 51.3% to 60.5%, while OpenAI's fell from 44.2% to 35.1%. Anthropic is taking a bigger slice — just as Bloomberg reports prospective IPO buyers discussing a $1.2–1.3 trillion valuation, down from about $2 trillion estimated in early October. The business itself is real: $17.0 billion in third-quarter revenue, $33.2 billion across nine months. The question is whether revenue growth keeps pace with the expectations baked into those numbers.

Two months is not a trend — but the assumptions are cracking

Before anyone declares the hypergrowth era dead: spending rebounded 8.9% from September 23 to October 23, and Ramp is clear its sample isn't representative of its total AI volume. The cause of the stall is still unresolved — thrift or price cuts, and in what proportion, nobody can yet say.

But Kharazian's line is the one to sit with: "You can use more AI, but use it for less. That's great for the customers… but it's not necessarily great for the model companies." The industry's hope is that personal agents restart the spending engine. Whether they'll be as economically productive as coding agents is, for now, an open question — and the market's favorite assumption just got its first serious audit.