Topic hub
Data Center Power Explained: MW, PUE, Grid and Cooling
Power is the scarcest input in the AI infrastructure build-out. This hub explains the terms used to describe data center power and why grid access now shapes where AI capacity is built.
Megawatts and capacity
Data center capacity is usually quoted in megawatts (MW) of IT load: the power available to servers, storage and networking. Campus announcements may quote total or planned capacity across several phases, so check what a figure refers to.
PUE: power usage effectiveness
PUE is total facility power divided by IT power. A PUE of 1.0 would mean every watt goes to IT equipment; real facilities add cooling, power conversion and lighting overhead, so PUE is always above 1.0. Lower is more efficient.
Grid constraints and energy procurement
New large loads require grid connections and transmission upgrades that can take years. Operators respond with long-term power purchase agreements, on-site generation, siting near available generation, and phased builds.
Cooling
Cooling is the largest overhead after IT load. Dense GPU racks increasingly use liquid cooling, which can move heat more efficiently than air at high densities.
Frequently asked questions
What does PUE mean?
Power usage effectiveness: total facility energy divided by the energy used by IT equipment. It measures overhead from cooling and power distribution.
How can I estimate the power cost of a GPU cluster?
Multiply GPU count by watts per GPU, apply PUE for facility overhead, then multiply by hours and your electricity price. Our data center power calculator does this for you.
Cite this page
“Data Center Power Explained: MW, PUE, Grid and Cooling.” GPU Data Hub. https://gpudatahub.com/topics/data-center-power
You are welcome to reference and link to this page. Please link to the URL above.