Amazon has flagged multiple AI technology projects with spending overruns, according to the Financial Times, adding a cautionary data point to the narrative that Big Tech’s artificial intelligence arms race will pay for itself sooner rather than later.
The company, which has committed to roughly $200 billion in capital expenditures for 2026 focused on AI infrastructure, is discovering what happens when you hand powerful tools to tens of thousands of employees and say “go build.”
The AI tool overuse problem
Internal reports revealed that Amazon employees were overutilizing AI tools for non-essential tasks. The result was a substantial spike in token consumption, the unit of measurement for how much compute power AI models chew through, and with it, rising operational costs that leadership didn’t anticipate.
Amazon implemented an internal AI usage leaderboard called KiroRank. The leaderboard was later removed after concerns about gamification surfaced, with employees competing to rack up AI usage stats rather than focusing on whether that usage was actually productive.
