Microsoft and Uber review AI spending as token costs increase

AI Cost Pressure Grows In 2026

Artificial intelligence promised that it would help companies reduce expenses and improve productivity. But in 2026, many firms are facing a different challenge. AI tools are becoming costly to use at scale, while clear productivity gains are still difficult to measure.

Microsoft has reportedly asked thousands of engineers to reduce the use of Claude Code and move to an internal AI tool by June 30. The decision is said to be linked to high token usage and rising costs. Uber has also faced a similar issue after giving Claude Code access to around 5,000 engineers in January. The company reportedly used its full annual AI budget within just five months.

Token Pricing Puts AI ROI Under Review

The main concern for companies is token-based pricing. AI tools charge based on input, output, prompt complexity, and usage volume. When thousands of employees use these tools daily, spending can rise quickly even if individual token prices fall.

Industry experts say this does not mean companies will stop using AI. Instead, businesses may become more careful about where and how AI is used. The focus is now shifting from adoption numbers to real business value. Companies are expected to measure whether AI tools are helping teams ship useful work, save time, or improve outcomes.

The next phase of enterprise AI may depend on controlled usage, better planning, and clear return on investment. AI is still important for businesses, but unmanaged use is becoming harder to justify.