Ask three suppliers for a 10 kVA UPS and you can receive three machines with three different real capacities. Power factor explains the gap. It is the ratio between apparent power, measured in kVA, and real power, measured in kW, and it remains the most common source of confusion and of undersizing in UPS procurement.

Do not compare kVA to kVA. Compare kW to kW, at the output power factor stated on the datasheet.

The old discount

Apparent power multiplied by power factor gives real power. A UPS rated 10 kVA at output power factor 0.9 therefore delivers at most 9 kW. At 0.8 it delivers 8 kW. For decades that derating was normal, because inverters and their output transformers could not push full real power across typical loads. Buyers who compared offers on the kVA figure alone bought less capacity than they believed they were getting.

What changed

Transformerless double-conversion UPS in the current generation carry a unity rating of PF 1.0, so a 10 kVA unit delivers a full 10 kW. Advances in inverter semiconductors, control firmware and thermal design produced that shift. The loads moved in the same direction: server and network power supplies with active power factor correction now present a power factor at or near 1.0, so the old assumption that a load draws only 0.8 of its apparent power no longer holds either.

Check the model, not the vintage

Unity rating is common now, but it is not universal, and assuming it by age of design will catch you out. Riello Sentryum and Salicru SLC ADAPT2 are rated at unity across their ranges. The Legrand Keor HP range, which is current and widely specified, is rated at 0.9: the 60 kVA frame delivers 54 kW, the 200 kVA frame delivers 180 kW. Both are sound machines. They are simply rated differently, and the datasheet is the only place that tells you which you are being offered.

Why it matters in practice

Capacity. At unity you can serve the same load from a smaller and cheaper frame, or keep the frame you planned and gain real headroom inside the same footprint.

Comparison. Two 20 kVA quotes are not equivalent when one delivers 20 kW and the other delivers 18 kW. That is a 10% difference in what you can actually connect.

Battery autonomy. Runtime is calculated from real power draw in kW, not from the kVA nameplate. A misread power factor distorts your expected runtime as well as your usable capacity, and it usually does so in the same unhelpful direction.

Growth. Sizing on kW with a stated margin gives a predictable expansion path. We generally recommend loading a UPS to around 75% of its real capacity, which leaves room for load growth without pushing the machine into its overload band.

Where generators complicate the arithmetic

Generator sets follow a different convention: they are rated in kVA at 0.8 power factor, so a 100 kVA set is an 80 kW machine. When a generator backs up a unity-rated UPS, comparing the two nameplates directly will undersize the set. Compare the real kW the UPS will draw against the generator kW rating, then add margin for the rectifier input characteristics, the battery recharge current and the step load at transfer. A naive kVA-to-kVA match almost always produces a generator that is too small.

How to buy correctly

Read the output power factor line on the datasheet before anything else, then compare kW to kW at that stated figure. Check the overload curve and the permitted leading and lagging load range, because unity-rated machines tolerate the capacitive IT loads that older designs handle poorly. Check the derating for altitude and ambient temperature, which reduces the kW you can use in a hot plant room in an Amman summer. And remember that legacy equipment in the field still carries the old ratings: replacing a ten-year-old 10 kVA unit rated at 0.9 with a modern 10 kVA machine gives you 11% more real capacity at no additional cost.

At Critical Energy Technologies we size every system from the measured load, the required runtime and the site conditions, never from a brochure figure. Send us your load list and we will return a properly engineered recommendation.