Nvidia says AI data centers waste up to 30 percent of power
Nvidia claims at a recent event that only 70 percent of electricity in AI data centers is used for actual processing, with the rest lost to inefficiencies.
Power usage breakdown details
Rod Evans, Nvidia's Vice President of Supercomputing and AI Cloud Infrastructure, stated that infrastructure and system issues account for the remaining 30 percent of power draw. Data center builder Vertiv shared this view during its annual EMEA event held in early 2026. The two companies have been collaborating since early 2026 to design sustainable facilities that address these losses.
Sources of energy loss
Inefficiencies include cooling systems, power distribution networks, and frequent hardware failures or restarts. These factors prevent the full utilization of every watt drawn from the grid for artificial intelligence tasks. Optimized future designs aim to reduce this non-processing consumption significantly.
Shift in capacity bottlenecks
While a chip shortage was previously seen as the main barrier to expanding AI data center capacity, grid connections are now the primary constraint. Vertiv provides power and infrastructure management for Nvidia's GPUs while working on these efficiency improvements.
Reported by one outlet
Only one outlet has published this. Nothing here has been checked against a second report, so read it as that outlet's account and follow the link below for the original.
Reported by
1 independent outlet. Headline as published. Links open the original report.