TechCrunch: Inside the industry scramble to manage AI's runaway costs

Rebecca Bellan’s June 5 piece in TechCrunch documents a pattern that is becoming common across engineering organizations: companies setting an annual AI tools budget at the start of 2026, then watching it evaporate months ahead of schedule.

Uber ran through its entire 2026 AI coding budget by April. Microsoft revoked Claude Code licenses from its developers months after enabling them. A Priceline employee told TechCrunch that a routine Cursor contract renewal came back four to five times more expensive than the year before. Per-developer token consumption rose approximately 18.6 times in nine months, according to research cited in the piece.

The numbers are not simply the result of waste. Engineers who used the most tokens were twice as productive — but consumed ten times more tokens to achieve those results. That trade-off is harder to justify than it sounds when token costs have risen alongside consumption. Goldman Sachs projects global token usage will multiply twenty-four times by 2030, which means cost pressure will compound for years. In response, the Linux Foundation launched the Tokenomics Foundation, an initiative to establish shared definitions and metrics across vendors so companies can actually compare what they are spending.

For product managers, this story signals a shift in how AI tool access is governed inside organizations. The early phase of “enable broad access and see what happens” is ending. Teams are now being asked to justify token spend the way they justify SaaS subscriptions — with productivity data, ROI estimates, and usage policies. PMs who oversee AI-assisted workflows or own tooling decisions will increasingly find themselves in conversations about budget control, not just feature development. Understanding basic token economics — what drives consumption, which tasks are cost-efficient, where agent loops create runaway spending — is becoming a practical skill rather than an optional one.