Companies Cut Back AI Spending As Costs Skyrocket

Several major technology companies are re-evaluating their extensive use of artificial intelligence, driven by mounting expenses. Amazon recently discontinued its internal AI leaderboard, Kirorank, which was intended to encourage AI adoption. The decision stemmed from two primary concerns: the significant increase in AI operational costs due to higher token usage and employees engaging in 'tokenmaxxing' to artificially inflate their scores on the leaderboard by assigning AI menial tasks. This practice led to substantial, unwarranted expenditures.

This trend is mirrored across the industry. Meta also ended an employee-run AI leaderboard due to similar tokenmaxxing issues. Uber has indicated struggles in justifying ongoing AI expenditures after rapidly consuming its projected 2026 budget within a single quarter. Microsoft has canceled certain AI licenses, and companies like Salesforce and DoorDash are reportedly rationing AI resources amid diminishing returns on investment and soaring costs.

Despite these corporate pullbacks, the overall consumption of generative AI remains at record levels. Google's Gemini platform, for instance, has seen a dramatic increase in token usage from 480 trillion in May 2025 to 3.2 quadrillion by May 2026. This surge is partly attributed to the growing popularity of agentic AI and coding tools that consume considerably more tokens than standard chatbot interactions. Experts suggest that while companies are beginning to differentiate valuable AI applications from mere usage, the demand for AI is expected to continue its upward trajectory, emphasizing the need for education on effective AI implementation and the reinforcement of human skills like critical thinking.