Rack Power Density Is Doubling Fast, and Data Centers Need to Catch Up
Average rack power draw jumped from 12 kW to 26 kW in just two years as AI workloads push facilities toward new limits on power delivery, cooling, and structural design.
All posts in this stream, newest first.
Average rack power draw jumped from 12 kW to 26 kW in just two years as AI workloads push facilities toward new limits on power delivery, cooling, and structural design.
Abandoned cloud instances, orphaned volumes, and idle GPU jobs are costing companies more than ever, and the rise of AI workloads is making them harder to find and kill.
The Department of Energy declined to designate three proposed National Interest Electric Transmission Corridors, even though its own draft study flags data centers as a major driver of future transmission needs, reviving debate over building grid capacity ahead of demand.
Despite years of awareness about IPv4 exhaustion, most operators still lean on CGNAT and dual-stack workarounds. A new industry analysis argues that agentic AI, edge computing, and cloud-native workloads are turning IPv6 from a nice-to-have into a strategic requirement.
From Meta's 5 GW Louisiana supercluster to new state restrictions in New York and Nebraska, August's data center news shows AI infrastructure racing ahead even as regulators start applying brakes.
Fujitsu is exiting datacenter ownership in Australia to focus on AI, cyber resilience, and quantum computing services, offloading five facilities to private equity firm Next Capital for a reported AUD$200 million.
Qualcomm unveiled a multigenerational CPU agreement with Meta and a new AI inference architecture, part of a broader push to build a $15 billion data center business by fiscal 2029.
A new ICF report finds US electricity demand could jump 39% by 2035, but the sharper constraint is whether transmission networks can actually deliver that power to fast-growing data center hubs.
The $500 billion Stargate initiative has expanded across the US and abroad, but energy constraints, financing structure, and local policy pushback are testing how fast the AI infrastructure project can actually scale.