Tool 002 · AI facilities

Turn a GPU count into a power plan.

Estimate average IT load, facility load at your PUE, heat rejection, cooling equivalent, energy, electricity cost, and rack density.

StatusFree to use · browser-onlyEvery Run Any AI resource declares what is real today.

Current estimate

Power, cooling, and energy plan

Use nameplate and measured-load assumptions you can defend. The calculator does not fetch vendor prices.

IT equipment
Facility and operating assumptions
Formula usedIT kW = GPU load + host overhead + network/storagefacility kW = IT kW × PUE

Energy uses 365.25 ÷ 12 days per average month. Heat conversion uses 3,412.142 BTU/h per kW and 12,000 BTU/h per cooling ton.

Read the result

Separate IT load from facility load.

IT load covers the modeled GPUs, host overhead, network, and storage. Multiplying by PUE adds the facility energy used to support that IT equipment. The difference is shown as cooling and facility overhead.

Use this as an early planning brief. Electrical protection, redundancy, transient loads, liquid-cooling equipment, and local code require qualified engineering.

Should I use TDP or measured power?

Use the best defensible average for energy planning and keep nameplate values for electrical worst-case review. The utilization field provides a simple average-load adjustment.

What does PUE include?

PUE is total facility energy divided by IT equipment energy. It captures cooling and other facility overhead as one ratio, but it does not explain which subsystem consumes it.

Why convert load to BTU/h and cooling tons?

Nearly all electrical power entering running equipment becomes heat. The conversions make early conversations with facility and cooling teams easier.

Is the cost a data-center quote?

No. It multiplies modeled kWh by your entered energy price. Demand charges, taxes, power contracts, capital cost, staff, and maintenance are outside the result.