Companies that once rushed to flood their operations with AI are now scrambling to rein in runaway token costs, as the bill for the AI gold rush comes due. TechCrunch reports that Uber burned through its entire 2026 AI coding budget by April, while Microsoft pulled developers' Claude Code licenses just months after enabling them. A Priceline employee described a routine Cursor contract renewal that came back four to five times more expensive than the previous year. The shift reflects a growing realization across corporate America: even though per-token prices have continued falling, the push for broader AI adoption and increasingly autonomous agents has driven consumption to unsustainable levels.
The new reality was on full display at a New York City event this week, where Alexander Embiricos, OpenAI's head of enterprise, told TechCrunch the conversation with customers has fundamentally changed. Six months ago, he said, buyers wanted to know what AI could do and whether it was good enough. Now the questions are about visibility, auditability, token controls, and model efficiency. That anxiety is what drove the Linux Foundation to announce the Tokenomics Foundation, a new standards body aimed at applying the same cost discipline to AI spending that the FinOps movement brought to cloud computing. J.R. Storment, executive director of the FinOps Foundation, said the shift in tone was dramatic: starting in April and May, he began hearing from companies that were three times over their entire 2026 token budget with months still left in the year. The conversation moved from "tokenmaxxing" and moving fast to "we need guardrails."
The strain traces back to late 2025, when CEOs pushed their teams to use the best available models regardless of price. The November release wave, including Anthropic's Claude Opus 4.5, OpenAI's GPT-5.1, and Google's Gemini 3 Pro, brought significant improvements to agentic tools capable of operating independently, multiplying token consumption far beyond simple chat use cases. The result is an emerging market of startups, established vendors, and standards bodies all racing to give enterprises the tooling and language to track, attribute, and control AI spend. Whether that infrastructure arrives fast enough to salvage ROI from the wreckage of overextended budgets remains the open question for an industry now collectively discovering that unlimited AI appetite eventually meets an unlimited invoice.