How much power does a typical AI data center use?

It varies enormously with scale — a single GPU rack can draw 40-130+ kW, a mid-size AI data hall can draw tens of megawatts, and the largest hyperscale AI campuses are being designed for several hundred megawatts to over a gigawatt of total IT load. The number that matters is total electrical capacity, not building square footage.

Why does an AI data center need so much more power than a traditional data center?

The increase comes almost entirely from GPU power density, not facility size. Modern GPU racks draw far more power per rack than traditional server racks did, so a facility can be completely full from a power-delivery standpoint while still having plenty of physical floor space left.

What is the difference between IT load and total facility power in a data center?

IT load is the power that actually reaches the compute equipment (GPUs, servers, networking). Total facility power is higher, because cooling, power conversion losses, and other supporting infrastructure also draw power. The ratio between the two is captured by the PUE metric — a facility with 100 MW of IT load and a PUE of 1.2 draws 120 MW total from the utility.

Why has AI data center power demand kept increasing?

Each new generation of AI accelerator hardware has generally drawn more power per chip, and the industry has largely chosen to deploy more and denser compute rather than hold total facility power flat. This rising trajectory is a major reason utility interconnection capacity has become the binding constraint on new AI data center construction.

Engineering·8 min read·September 19, 2026

🔋 How Much Power Does an AI Data Center Need?

A grounded look at the actual power numbers behind an AI data center — from single-rack draw through hall-scale and campus-scale totals — and why total IT load, not building square footage, is the number that actually drives design.

🎓Data Center Engineering Basics — Start to Finish (Full Access)

View →

Power, Not Floor Space, Is the Real Design Constraint

For a traditional data center, usable capacity was often described in square footage or the number of racks a building could hold. For an AI data center, the binding constraint is almost always electrical power — a facility can have plenty of physical floor space left and still be completely full from a power-delivery standpoint, because GPU racks draw so much more power per square foot than traditional servers ever did. This is why AI data center capacity is described in megawatts of IT load, not square feet, and why utility interconnection capacity (not land or building size) has become the real bottleneck on how much a given site can host.

Topics covered

how much power does an AI data center needAI data center power consumptionAI data center MWdata center power requirementshow many megawatts does a data center useAI data center energy useGPU data center power consumption