Just 1% of companies account for 80% of all business spending on OpenAI and Anthropic, according to corporate card giant Ramp, which monitors payments across more than 70,000 U.S. businesses. Lead economist Ara Kharazian, who flagged the concentration in early September, noted that the spending pattern is more concentrated than any other software category the firm tracks. The report reveals that despite rising token volumes, total dollar spending on AI has essentially flatlined, with the top 1% of spenders shelling out less per employee in September than they did in July.

The spending breakdown shows venture-backed startups dominate AI expenditures. Ramp's summer data showed the median VC-backed startup spent $76 per employee monthly on AI, more than ten times the $7 spent at private equity-backed firms. Among the heaviest spenders, the top 1% of AI-spending companies paid out $9,568 per employee in September, up 11.7% from August but still 3.1% below July's peak of $9,873. At Anthropic specifically, the top 1% of customers generated 46% of revenue in August, nearly double the 25% share from a year earlier. Just two unnamed customers produced nearly a quarter of Anthropic's 2025 revenue, with AI coding tools Cursor and Microsoft's GitHub Copilot driving roughly $1.2 billion of the lab's $5 billion annualized revenue run-rate last year. Across Ramp's corporate network, paid AI adoption edged up only 0.62 percentage points to 55.9% in September.

In his October AI Index—titled "Cracks in the AI Thesis Part 3"—Kharazian reported that August and September recorded "two straight months of virtually no growth in token spend," even as raw token volumes continued climbing. Total dollar spending did reach a new high in early October, but cleared its previous August peak by just 1.5%. According to Kharazian, the main culprit is deflation: "the main driver is price declines at the frontier, in addition to efficiency and quality improvements for models across intelligence levels, that are driving the blended price per million tokens down." Kharazian wrote that "it feels like we're at the top of the S-curve, where the growth in the market will have to come from more intense adoption rather than more customers."

The price collapse is squeezing margins throughout the AI supply chain. Anthropic's realized token price has been slashed in half in just a month, plummeting to $0.83 per million tokens from over $1.70 in mid-September—a record low since the Ornn Anthropic Token Price Index launched in June. OpenAI's realized pricing sits even lower at $0.45, after the company cut rates on its GPT-5.6 Luna and Terra models by up to 80% in late July. Goldman Sachs warned in a September 25 note that "token demand growth will need to outpace declining token prices to support continued growth in investment spending," noting that while total token demand has surged roughly 18-fold since December, demand for high-margin frontier models has grown only half as fast. The investment bank calculated that realized return on invested capital across Amazon, Alphabet, Meta, Microsoft, and Oracle has tumbled from 40% to 25%—the lowest reading in its dataset going back to 2016. Apollo chief economist Torsten Slok concluded that "the number-one variable governing the viability of data center construction models is not physical capacity or power; it is the actual price users are willing to pay for the compute being built." The concentration of spending among a tiny fraction of customers—many funded by the same venture firms backing the AI labs themselves—suggests the AI revenue boom may be more circular than sustainable. If enterprise buyers won't lock in long-term contracts at premium prices, and if adoption has already plateaued near 56%, the path to profitability narrows considerably for labs burning billions on compute.