Companies Are Hitting AI Budget Walls. A New Market Is Racing to Fix It.
Tech companies are burning through AI spending faster than their budgets allow. Uber exhausted its entire 2026 AI coding budget by April. Microsoft revoked developer access to Claude Code licenses after enabling them. Priceline saw a routine Cursor contract renewal jump 4-5x in price.
The problem isn't token prices - they've fallen. The problem is consumption. As companies pushed harder on AI adoption and deployed increasingly autonomous agents, token usage exploded. Teams that signed up for all-you-can-eat subscriptions in early 2025 are now scrambling to understand where money is going and whether they're getting any return.
One company reportedly racked up a $500 million Claude bill after forgetting to set usage limits.
The Conversation Has Shifted
Alexander Embiricos, OpenAI's head of enterprise, said the shift happened fast. "Six months ago, conversations were about whether AI was good enough," he said. "Now they're about visibility, auditability, token controls, and model efficiency."
The Linux Foundation this week announced the Tokenomics Foundation, a standards body aimed at bringing cost discipline to AI spending the way FinOps did for cloud costs.
J.R. Storment, executive director of the FinOps Foundation, said the crisis hit in April and May. "Companies told us they were 3x over their entire 2026 token budget," he said. "The conversation shifted from 'go fast' to 'we need guardrails.'"
Productivity Gains Don't Match the Spending
A study of 20,000 developers found engineers using the most tokens were roughly twice as productive as those using AI less - but they spent 10 times more tokens to achieve it.
Per-developer token consumption rose 18.6x in nine months, according to engineering management platform Jellyfish. That spending surge came largely from agentic features released in November by Anthropic, OpenAI, and Google.
The real problem: most companies can't measure whether extreme spending actually produces business value. "Whether extreme spend pays off comes down to the revenue generated from shipped code," said Nicholas Arcolano, head of research at Jellyfish. "Most companies still can't measure that."
The Scale Problem
Tracking token costs is fundamentally different from tracking cloud costs. Cloud cost tracking handles hundreds of millions of rows per month. Token tracking requires handling trillions of rows per month.
At Priceline, finance teams are already spotting discrepancies between vendor reports and internal data. Chris Reed, senior director of IT finance at Priceline, compared it to telecom expense management. "Anytime you introduce something new, it's ripe for billing errors and optimization opportunities," he said.
A Market Is Forming
Pure-play vendors like Pay-i and Paid are building tools to track and optimize AI spending. Engineering platforms like Jellyfish, Waydev, and Faros AI are adding AI agent monitoring to prove ROI. Established companies are moving in too: Ramp launched AI spend management, while Datadog and New Relic added token-level observability and GPU monitoring.
AWS is expected to announce new financial management features for enterprise AI spending at the FinOps X conference next week.
Model providers themselves are optimizing. Some are automatically routing queries to cheaper models - so an Anthropic bill for Claude Opus may include spend on Sonnet or Haiku. "I think this will become more and more of a thing," said Vitaly Gordon, CEO of Faros AI.
The Missing Piece: Standards
All these tools are being built without shared definitions for token costs, output value, or cross-vendor comparisons. The Tokenomics Foundation aims to fill that gap with canonical definitions, open standards, and new metrics like cost-per-intelligence and tokens-per-watt.
The foundation plans a formal launch in July. More members will be announced at FinOps X next week.
"Token economics is fundamentally more abstract and opaque than anything we've managed at this scale before," said Nishant Gupta, chief availability officer at Salesforce. "It requires a different operational muscle than the one the industry built for cloud."
Goldman Sachs projects global token usage will multiply 24 times by 2030. Companies already over budget need solutions now, and the foundation's first deliverable is still months away.
The Smart Play
The best ROI doesn't come from pushing heavy users higher. It comes from moving the broad middle from low to moderate usage, according to Arcolano. Most companies would benefit from moderate, broad adoption rather than concentrated spending by a few power users.
For managers overseeing AI adoption, the message is clear: control is becoming as important as capability. Set usage limits. Track spending against business outcomes. Don't assume more tokens means more value.
Your membership also unlocks:
