ChipsThe story, in brief
Just how much power will AI need?
Agentic AI could redraw the compute map: data centers aren't ready for the power demands ahead.
The KeyNews take
Why it matters
As AI shifts from inference to multi-step autonomous reasoning, power consumption per workload will spike — forcing a reckoning on data-center capacity, cooling, and grid infrastructure that practitioners and operators need to plan for now.
The key facts
4 to knowArticle examines power scaling as agentic AI workloads replace single-inference tasks
Focus on data-center electricity demand and infrastructure readiness
Implications for compute economics and on-prem vs. cloud strategy
Published Sep 2026 — positions this as an emerging constraint for builders and operators
Go to the source
Financial Times Technologyft.com
Publisher excerpt: Rise of agentic AI could supercharge the electricity demands of data centres