Both Microsoft and Uber are hitting the brakes; enterprise AI deployment has finally moved from the "fear of missing out" stage to the "careful calculation" stage, and large-scale scenarios beyond programming will have to wait a bit longer.

Mars Finance reports that after large-scale enterprise deployment of AI, costs and returns have become focal points. Microsoft has canceled most Claude Code licenses, and Uber's COO also stated that AI expenses are difficult to justify. An AI consultant said that clients who do not restrict employee use of Claude could spend up to $500 million per month. Industry experts warn that token maxxing and full AI integration often yield poor ROI, with truly mature large-scale scenarios mostly in programming. Blind expansion leads to rising IT costs, and even querying simple questions like weather with high-cost models is not unlimited. Companies are offsetting AI bill pressures through layoffs, claiming this is a possible way to cover costs.