Enterprise AI Spending Hits a Wall as ROI Remains Elusive

The AI spending reckoning arrives

A pattern is emerging across enterprise technology: organizations that rushed to deploy AI are now confronting costs they can't justify. Uber's COO recently acknowledged that AI expenses are becoming "harder to justify." Microsoft has canceled most of its Claude Code licenses over spending concerns. One organization reportedly burned through $500 million in a single month after failing to implement usage controls on employee AI licenses, according to Axios.

The culprit, according to IBM's newsroom, is what some are calling "tokenmaxxing"—the organizational imperative to use as much AI as possible, as quickly as possible. For two years, competitive pressure drove adoption faster than financial discipline. Without meaningful metrics, companies created usage leaderboards that employees learned to game, turning consumption into a false proxy for value.

Now the bills are arriving, and chief financial officers are demanding answers.

Why it matters

This isn't a story about AI failing to deliver—it's about enterprises lacking the financial frameworks to measure what they're getting. Without the ability to tie AI spending to business outcomes, even successful implementations become vulnerable when budgets tighten. Organizations that can't demonstrate ROI will struggle to maintain AI investments regardless of their strategic value.

The measurement gap

Most enterprises can track what they pay for AI licenses, but few have reliable systems for connecting token consumption and compute costs to actual business results. For the past two years, AI has been treated as a technology initiative rather than a business transformation, with little scrutiny of whether returns were appearing on profit-and-loss statements.

Research from the IBM Institute for Business Value reveals the scale of this disconnect: 79 percent of executives expect AI to drive significant revenue by 2030, yet only 24 percent can identify where that revenue will originate. Without defined outcomes, measuring returns becomes impossible—and defending spending to boards becomes untenable.

What disciplined adoption requires

The solution isn't cutting AI budgets wholesale, but focusing them strategically. When AI is deployed as a general-purpose productivity tool layered onto existing workflows, token spending has no anchor. When embedded in specific processes tied to measurable outcomes, the economics shift fundamentally.

Every AI use case should connect to a defined workflow and measurable outcome, with returns tracked in three-to-six-month increments. Like any capital allocation decision, if an AI investment can't deliver 2.5x to 3x returns—whether through time savings, improved customer and employee experiences, or new revenue—it shouldn't proceed.

The organizations that will succeed with AI aren't necessarily those that adopted earliest or spent most aggressively. They're the ones building the financial discipline to measure what matters and the strategic clarity to focus resources where returns are demonstrable.

These details were first reported by IBM's newsroom in a post by Neil Dhar.