Gen AI power consumption set to double in 2026

In May 2025, Uptime Intelligence noted that data center executives had recognized a disconnect between infrastructure investments and end-user revenues from applications built on generative AI models, raising questions about the durability of the current trend for very large, high-density facilities (see Gen AI power consumption surges higher faster). Since then, that question has only grown larger.

Nvidia’s latest quarterly earnings and outlook outstripped already elevated expectations, defying concerns about high spending and unclear returns on generative AI compute. Surging shipments of data center GPUs and associated system components — from already historic volumes — send the signal that large customers (hyperscalers, neoclouds and AI developers) are following through on their investment commitments to build immense AI compute capacities.