How much power does an H100 GPU use?

Planning often uses 300–700 W per accelerator depending on SKU, workload, and cooling; sum GPU + CPU + network + cooling overhead for rack totals.

Power calculator hub → Scenario

GPU Rack Power for AI Infrastructure

Q: How do I size UPS for a GPU rack?

Convert rack kW to kVA with PF, add margin, then check runtime or use redundant UPS modules for N+1 paths.

Q: Does liquid cooling change rack kW?

IT kW is similar; facility kW may drop PUE because heat is rejected more efficiently—still size electrical feed for full rack demand.

Screen kW per GPU, kW per rack, and row-level demand for AI training or inference clusters before transformer, UPS, and liquid-cooling design.

Who this is for

Best for: Teams planning AI rack power density (GPU count × TDP + host + network overhead).

Not ideal for: Network fabric architecture or chip-level thermal CFD.

Quick answer

Rack kW ≈ (GPU count × GPU kW) + host kW + network kW + **15% margin**. Per-GPU planning often uses **0.3–0.7 kW** for datacenter accelerators—verify OEM TDP for your SKU.

Roll up rack kW

Example: 8× GPU rack

8 GPUs × 0.5 kW = 4 kW; host + NIC ≈ 2 kW; subtotal 6 kW; 15% margin → **~6.9 kW** rack planning figure. At PF 0.95 → **~7.3 kVA** for UPS branch screening.

Liquid cooling note

Direct-to-chip liquid cooling may improve PUE but **does not reduce electrical feed kW** at the rack PDU—size power for full IT draw. See PUE planning for facility overhead.

Next steps

Disclaimer

GPU TDP varies by workload and vendor. Use metered data and OEM specs for procurement and electrical design.

FAQ

How much power does an H100 use?

Planning range ~0.3–0.7 kW per GPU plus host/network; confirm SKU datasheets.

How do I size UPS for a GPU rack?

kVA = kW ÷ PF with margin; check N+1 redundancy separately.

Does liquid cooling change rack kW?

IT kW stays similar; facility PUE may improve.