How fast do banks respond to AI cost spikes?

J.R. Storment, executive director of the Tokenomics Foundation, told Bloomberg that "during the early rush to cloud services more than a decade ago, banks took at least five years to control costs that rose with usage and bring them under a dedicated budget line." By contrast, he said, "what's different now is that financial chiefs are reacting to AI cost spikes within 6 to 18 months of adoption," adding that this speed is itself "making it difficult to put a final price tag on premium AI models."

Storment's comparison locates the distinguishing feature of this cost cycle in its speed rather than its size: cloud spending crept up gradually enough that enterprises had years to build governance before costs became unmanageable, whereas AI inference cost scales directly with usage intensity — more reasoning steps, more agentic turns, more tokens per task — so the same CFO alarm that took roughly five years to sound for cloud is sounding in under two years for AI. Bloomberg's broader reporting shows governance measures that took years to build for cloud being improvised in real time for AI: JPMorgan's CFO Jeremy Barnum cited an annual AI expense of about $700 million while acknowledging "a precise figure is hard to pin down because the technology is now spread across the entire bank," and some JPMorgan employees were reported to be "spending more on tokens than their salary" before spending limits existed.

This sharpens How fast are frontier AI training costs actually rising? by adding the deployment-side mirror of that training-side cost growth: it is not only frontier labs' training runs that are compounding fast, but ordinary enterprise inference bills, on a timescale that is outrunning existing budgeting practice. It also complicates How fast is AI capability becoming cheaper to produce?: falling per-token prices have not been enough to prevent the spikes Storment describes, because agentic workflows appear to be growing token volume per task faster than per-token prices are falling. And it supplies the timeline context for Can banks cut AI costs by routing tasks to cheaper models?, the routing and capping strategies banks built once cost spikes forced the issue.

Storment's comparison is one consultant's characterization relayed by Bloomberg, not a measured study of enterprise cost-control timelines in either era, and the excerpt does not specify what "control" meant in the cloud case beyond "a dedicated budget line" — a governance milestone, not evidence that cloud costs stopped rising. Held to that limit, the implication is that enterprises adopting agentic AI should expect cost governance to become urgent on a timescale of months rather than years, and should plan monitoring and routing infrastructure from the start rather than after a surprise bill forces the issue.