Particle.news
Download on the App Store

Companies Pull Back on Expensive AI Use as Uber Caps Employee Token Spending

Skyrocketing per‑token bills from agentic coding tools have prompted firms to impose per‑employee caps, deploy usage dashboards, limit licenses, and demand greater model efficiency.

Overview

  • Uber set a $1,500 monthly token‑spend cap per employee for each agentic coding tool and gave staff dashboards to track usage while allowing exceptions with permission.
  • Uber’s leadership said the company exhausted its full‑year 2026 AI budget within months after earlier encouraging broad internal use and running internal leaderboards that boosted consumption.
  • OpenAI CEO Sam Altman said his top internal user consumes about 100 billion tokens a month and that customers have started calling token costs a “huge issue,” pressing providers to cut waste and improve efficiency.
  • Other large firms including Microsoft, Amazon and Walmart have reduced third‑party licenses or restricted employee access to agentic tools as executives question whether heavy token spending delivers measurable product gains.
  • Tokens are the unit LLMs process and agentic agents burn far more tokens than simple chat, so rising per‑token bills are already affecting hiring plans, procurement rules and how companies judge AI’s ROI.