The narrative that artificial intelligence will generate limitless demand has encountered a significant obstacle, according to a post published October 20, 2026 by Kurt S. Altrichter, CRPS®, citing data from the corporate spend management platform Ramp. Corporate token spending growth plummeted from 42% in July to just 0.2% in September, while the five largest hyperscale cloud providers saw their return on invested capital cut nearly in half as capital expenditures surge faster than profits can keep pace.
The spending data reveals a dramatic deceleration in how quickly businesses are ramping up their use of AI services. In July, month-over-month growth in corporate token spending stood at 42%, reflecting rapid adoption and scaling of AI tools across enterprises. By September, that growth rate had essentially flatlined at 0.2%, suggesting demand may be stabilizing or hitting natural limits far sooner than many projections anticipated. At the same time, return on invested capital at the five biggest hyperscalers dropped from 38% to 20%, indicating that the massive infrastructure investments required to support AI workloads are outpacing the revenue and earnings those investments generate.
According to Altrichter, "The infinite-demand thesis needs infinite returns. The returns are going the other way." The post characterizes the situation as the moment when the "unlimited AI demand" story encountered a barrier, pointing to the stark mismatch between capital spending and the financial returns those investments are producing.
The underlying issue centers on a fundamental tension in the AI infrastructure buildout: cloud providers have been pouring capital into data centers, specialized chips, and network capacity based on forecasts of exponential demand growth, but actual customer spending appears to be leveling off far earlier than those models predicted. When return on invested capital falls by nearly half in a matter of months, it signals that each dollar of new infrastructure is generating significantly less profit than previous investments did. If corporate customers are no longer accelerating their AI usage at anything close to July's pace, the economics of continued aggressive expansion become harder to justify. The gap between what companies are willing to spend on AI services and what providers need to earn back on their infrastructure investments may be widening rather than closing, raising questions about whether the current investment cycle is sustainable at its present scale.
The data suggests the AI investment wave may be entering a new phase where growth stabilizes and providers must demonstrate profitability rather than simply scale. If spending growth remains near zero in coming months, hyperscalers may need to recalibrate their capital plans or find new revenue streams to justify the infrastructure already deployed. The takeaway is stark: betting on infinite demand requires infinite returns, and the returns are moving in the opposite direction. For executives weighing major AI commitments, the calculus has shifted from "how fast can we scale" to "what return will we actually see," and the early answer from the market's largest players offers a sobering reality check on both sides of the transaction.