For High-Density AI, Available Data Center Space May Not Be Usable
North American vacancy remains at 1%, but JLL says only a very small share of that capacity can support high-density AI deployments.
Get smart on it
Data centers report only 1% vacancy in North America, but most of that available space cannot actually support the extreme power densities required for advanced AI systems. A buyer seeking capacity for 140 to 160 kilowatts per rack found that nearly every operator had space available, yet almost none could deliver the necessary infrastructure, particularly direct liquid cooling systems. The gap exists because most data centers were built around much lower power requirements, and upgrading to support these extreme densities requires substantial changes to cooling, electrical distribution, and structural systems. Only a very small portion of the overall vacant capacity can meet the specialized technical requirements that high-density AI deployments demand.
Related news
Home-Based GPU Networks: Viable Supplements to AI Data Centers?
Distributed computing is getting a new spin. A growing crop of pilots is paying homeowners to host GPU capacity via wall-mounted appliances. Can residential nodes deliver the speed, reliability, security, and scale?
Same Cluster, 33 Points More Utilization: What Changed Was the Order
Reordering how tasks run on shared computing resources can significantly boost efficiency without requiring new hardware or infrastructure investment.
