AI Bills Are Arriving and Enterprises Are Not Ready for Them

Quick Facts

• 73% of enterprises reported AI costs exceeded original projections, according to the FinOps Foundation's 2026 State of FinOps report.
• Uber exhausted its entire 2026 AI budget by April after rolling out Claude Code to roughly 5,000 engineers in December 2025.
• Fewer than 10% of enterprises report measurable ROI on AI investments despite global enterprise AI spending crossing $400 billion, according to Draup.

The bills are in. And for most enterprises, they are larger than anyone expected.

AI adoption accelerated fast over the past two years. Boards pushed for deployment. Executive teams launched initiatives. Business units raced to find use cases. Now, technology leaders across industries are confronting budget overruns, runaway token consumption, and invoices they did not see coming.

The FinOps Foundation's 2026 State of FinOps report found 73% of enterprises reported AI costs exceeded original projections. The average enterprise AI budget grew from $1.2 million per year in 2024 to $7 million in 2026. KPMG's latest survey projects U.S. enterprises will spend an average of $207 million on AI over the next twelve months, nearly double the figure from a year ago.

The Uber Warning

Uber is the clearest case study. The company rolled out access to Claude Code to roughly 5,000 engineers in December 2025. Usage nearly doubled by February 2026. By March, 84% of developers were classified as agentic coding users. By April, the entire 2026 AI budget was gone.

Monthly API costs per engineer ran between $500 and $2,000. Per-developer token consumption rose 18.6 times in nine months. Uber's CTO said publicly he was back to the drawing board.

Uber is not alone. One company reportedly spent half a billion dollars in a single month after failing to set usage limits on its Claude licenses. Microsoft canceled most of its internal Claude Code licenses, partly over cost, six months after rolling them out.

The Visibility Problem

A KPMG Q2 2026 AI Pulse survey found only 26% of enterprises have real-time visibility into AI operating costs, even among companies with more than $1 billion in annual revenue. Another 42% have only partial visibility. And 62% of organizations still cannot accurately predict their monthly AI expenses.

Salesforce CEO Marc Benioff said his company's Anthropic bill will reach around $300 million this year. He said he wished there were a smart router that could determine which queries actually needed the most capable, most expensive models.

KPMG's Edwige Sacco warned that the growing practice of token-maxxing, where organizations gamify AI usage through internal leaderboards, risks incentivizing activity over outcomes.

ROI Is Elusive

Global enterprise AI investment has crossed $400 billion. Yet fewer than 10% of enterprises report measurable ROI, according to Draup. A Gartner survey of 506 CIOs found 72% reported their organizations are breaking even or losing money on AI investments.

MIT's 2025 GenAI Divide report found 95% of enterprise generative AI pilots fail to deliver measurable ROI or scale beyond experimentation. Gartner has predicted that more than 40% of agentic AI projects will be canceled by 2027.

IBM's Institute for Business Value found that computing costs are expected to climb 89% between 2023 and 2025, with 70% of executives citing generative AI as a critical driver. Every executive IBM surveyed reported canceling or postponing at least one generative AI initiative due to cost concerns.

The Billing Model Is Part of the Problem

Several major AI providers, including Anthropic and OpenAI, have moved services toward usage-based billing rather than flat-rate subscriptions. Token-based pricing ties cost directly to usage. Agentic workflows consume five to thirty times more tokens than simple chatbot interactions, making cost projections far harder to pin down.

Rahsaan Shears, AI Enterprise Transformation Leader at KPMG, said AI agents are changing both the operating model and the economics. The implication is clear: finance teams built their models around subscription costs. They were not ready for token bills.

Uber COO Andrew Macdonald said publicly it was very hard to draw a line between AI-assisted code commits and whether the company was actually shipping more useful features to consumers. That measurement gap is central to the problem. Productivity gains from AI often do not appear as credits on an invoice. The costs do.

The organizations that come out ahead will be the ones that establish real-time cost visibility, tie spending to measurable outputs, and set usage controls before deployment, not after the April bill arrives.

Read more: Don't count the savings until you know what the AI actually costs