TOOLS · TECHNICAL GUIDES · ELECTRICAL INFRASTRUCTUREWrite for us & partner ↗
Planning hub

AI compute needs a complete power model.

Build an explicit chain from equipment inputs to IT demand, facility overhead and capacity assumptions.

Interactive calculator · explicit assumptions

AI Data Center Power Calculator

Calculations run in your browser. Sharing puts numeric assumptions in the link. No registration required.

Three boundaries to keep separate

  1. GPU component: one part of the server, not the total AC input.
  2. IT equipment: servers, networking, storage and other IT loads within your boundary.
  3. Facility: IT plus cooling, electrical losses and other building overhead.

Use whole-server data or measurements where possible. If using chip power, explicitly add the rest of the system; do not call that value a complete server rating.

Carry inputs forward: use “Use IT load in capacity planner” below the calculator, then review and apply the IT MW and PUE on the next page. Other planning assumptions are preserved. You can also enter these values manually.

Continue to capacity sandbox →

Sources & further reading

  1. NVIDIA DGX H100 data center planning documentation ↗
  2. ENERGY STAR: energy boundaries and PUE ↗

Sources checked October 8, 2026. Follow the source for revisions and regional product details.

Go deeper

AI data centers

GPU TDP vs facility power

A worked example and the assumptions to document.

Explore resource →
Free calculator

Measured PUE vs planning PUE

Different uses for an energy ratio and a scenario multiplier.

Explore resource →
Free calculator

Explore the energy bill

Use load fraction and your tariff, without invented market prices.

Explore resource →