Nearly three-quarters of organizations have been hit with unexpected AI cost spikes over the past year as the "tokenmaxxing" trend sweeps through enterprises, mirroring the wasteful spending patterns of the early cloud era. A new study from Harness shows AI now accounts for 23% of the average enterprise cloud bill, with 26% of that spend wasted - and FinOps practices that tamed cloud costs a decade ago are now being pressed into service to bring AI bills under control.
Patrick Brogan, director of the FinOps advisory team at Harness, said the current wave of AI experimentation has companies "throwing everything at the wall to see what sticks." The result is invoice shock. Uber was a standout example: as reported in June, the ride-hailing firm burned through its entire annual AI budget in just four months after incentivizing staff to engage in tokenmaxxing.
"There's a number of factors that go into tokenmaxxing," Brogan said. "Some of it has to do with organizations deliberately pushing their employees to use the technology. It might come from a sort of subconscious desire, subconscious FOMO. They don't want to feel like their competitors are leveraging technology in a way that's going to put their own company at a disadvantage."
Ownership vacuum drives waste
Harness found that 52% of enterprises have no clear owner for AI costs, with responsibility fractured across engineering, finance, and IT. That governance gap, Brogan said, is fundamentally the same problem that plagued cloud spending a decade ago. "AI cost management has the same problems that cloud had," he said. "When we were getting started in our FinOps practice there, we felt the same pain points. You know, confusion over ownership, gaps in how we govern the cloud bill and the services that our organizations will use, and certainly invoice shock."
The core FinOps framework - which unites engineering, finance, and business teams to establish accountability for spending - is not a new discipline. But it has been sidelined during the AI boom. Brogan said companies need to "agree and align on a single owner of the AI bill" immediately, whether that's one person or an entire team. Without that clear decision, spending can ramp up unchecked in areas like software engineering, where no one function is designed to answer for it.
For managers looking to build these capabilities, AI for Management training can help teams understand the cost implications of different models and embed financial accountability into AI workflows.
Speed compresses the problem
While the cost management challenges echo the cloud era, the timeline is compressed. "Those problems are compressed into a fraction of the time because AI technology is developing at such a rapid pace and its usage and adoption have exploded far faster than cloud," Brogan said. "We don't have the same 10-year span to figure out the solutions to the challenges with AI spend but, luckily, we have a lot of lessons we learned with managing cloud spend."
Nearly six in ten engineers told Harness they are still encouraged to engage in tokenmaxxing, suggesting that organizations have not yet internalized the need for restraint. Some consultancies have taken drastic action: Accenture recently told staff to cut down on AI use for basic tasks due to "soaring token spend."
Model choice is a cost lever
A key error enterprises make is defaulting to expensive frontier models for tasks that could be handled by simpler, cheaper alternatives. Nitish Tyagi, senior principal analyst at Gartner, said "intelligent model routing" strategies are now a focus for developer teams. Brogan agreed, urging firms to select models that are "best fit for purpose."
"Not using a frontier model for something that can be delivered by an older generation. Maybe it arrives 10% slower at that outcome, but if you can live with that trade-off for the lower cost, then that should be built into the application design from the start," he said.
Why this matters for management
The AI cost crisis is a governance problem, not a technology problem. Managers who treat AI spending like a variable engineering expense without a clear owner will repeat the mistakes of the cloud's early days - only faster. Assign a single accountable owner for AI costs, demand cost visibility at the model level, and enforce a "fit for purpose" model selection policy that accepts slight performance trade-offs for significant savings. Without these steps, the 26% waste rate found by Harness will persist, and surprise bills will continue to hit the bottom line.
Your membership also unlocks:
