Size the UPS for a GPU cluster in seconds. Enter your IT load, power factor and redundancy target and get the required real power (kW), apparent power (kVA) and how many UPS units you need — with the modular UPS that fits.
A UPS keeps a GPU cluster alive through utility dips and the seconds before generators pick up. Size it too small and you lose the load; too large and you waste capital and efficiency. This tool turns an IT load and a redundancy target into the real power, apparent power and unit count you need.
Each step is a simple, labelled calculation so you can see exactly where the numbers come from and match them to your design.
GPU racks change UPS sizing in two ways. First, they are dense and near-continuous, so the protected load is large and the UPS runs close to its rating — efficiency and headroom matter. Second, they are mission-critical: a lost training run or dropped inference fleet is expensive, so most AI clusters run N+1 or 2N redundancy rather than bare N.
The key conversion is real power to apparent power. A UPS is rated in kVA, but your load is quoted in kW, and the two differ by the power factor: kVA = kW ÷ power factor. At a 0.9 power factor a 1,000 kW load already needs about 1,111 kVA before any redundancy. Apply the redundancy factor on top, divide by the unit rating and round up, and you have the number of UPS units — with the installed kVA telling you the real capacity on the floor.
Modular UPS systems like the 1,500 kVA unit this tool sizes against let you add capacity in blocks as the cluster grows and swap a module without dropping the load. Getting the sizing right early is what lets you lock UPS and switchgear — both long-lead items — before they gate the build.
The redundancy tiers the calculator applies and what each means for availability. Confirm the topology against your uptime target and utility/generator arrangement.
| Tier | Factor | What it means |
|---|---|---|
| N | 1× | Exactly enough capacity, no spare |
| N+1 | 1.2× | One extra module — survives a single failure |
| 2N | 2× | Two independent systems, each carries 100% |
Start from the IT load you need to protect, then apply two adjustments. First, convert real power (kW) to apparent power (kVA) using the power factor: kVA = kW ÷ power factor. Second, multiply by a redundancy factor (N, N+1 or 2N) so the UPS can carry the load with a module or string out of service. Divide the result by the unit rating and round up to get the number of UPS units.
GPU racks are dense and near-continuous, so UPS sizing tracks the full IT load. Required real power = IT load × redundancy factor. Apparent power (what the UPS is rated in) = IT load ÷ power factor × redundancy factor. This calculator sizes against a 1,500 kVA modular UPS, so unit count = ceil(required kVA ÷ 1,500).
kW is real (working) power; kVA is apparent power — the product of voltage and current the UPS must actually handle. They are related by the power factor: kW = kVA × power factor. A UPS is rated in kVA (and kW), so you size in kVA using kVA = kW ÷ power factor. At a 0.9 power factor, a 1,000 kW load needs about 1,111 kVA before redundancy.
N is exactly enough UPS capacity for the load with no spare. N+1 adds one extra module/unit so a single failure or maintenance does not drop the load — roughly a 1.2× sizing factor. 2N is two fully independent systems, each able to carry 100% of the load, for the highest availability — a 2× factor. Denser, mission-critical AI clusters typically run N+1 or 2N.
Our solutions engineers size UPS, distribution and cooling together and confirm real lead times — no payment, no commitment, quotes back in about one business day.
Our engineers scope power, cooling and compute together. No payment, and quotes come back in about a business day.