Tech giants impose AI usage limits as firms confront rising costs

Major corporations are curbing employee access to internal AI tools after budgets spiraled out of control. Walmart set a cap on usage of its AI assistant, Uber imposed a $1,500 per‑employee limit on code‑generation tools, and Microsoft disabled its own AI services for engineers. Amazon halted an internal leaderboard that rewarded token consumption, while Meta engineers burned through over 60 trillion tokens in a month, amounting to roughly $900 million in costs. Companies that previously linked bonuses to AI adoption, such as Starbucks and Accenture, are now scaling back in response to budget overruns – a Bloomberg‑cited 71 % of firms say they have exceeded their AI budgets for 2025.

At the same time, the surge in workplace AI use is straining existing hardware. Older PCs cannot handle the processing demands of AI features like background‑blur and real‑time transcription, prompting vendors to promote new AI‑optimized devices. Intel’s latest AI‑PC platform can deliver up to 180 TOPS, combining CPUs, NPUs and GPUs to offload AI workloads. Microsoft and Intel argue that locally running smaller models, such as PhiSilica, can improve speed, reduce power use and enhance data security, offering employees potentially five extra productive hours per week.