How to Size a UPS for a Server Room (Australian Guide 2026)

The short version

For a typical Australian server room, the rough sizing is: total connected wattage × 1.25 (headroom + inrush) ÷ 0.9 (power factor) = required UPS kVA. Then add the runtime you actually need.

That formula gets you within striking distance for 80% of single-phase server rooms. The remaining 20% — sites with non-linear loads, motor-driven equipment, or aggressive growth plans — need a more careful walk-through. The rest of this guide unpacks every variable.

Step 1 — Build a connected-load list

Walk every rack and every protected device and capture the nameplate wattage. For modern servers, the nameplate is usually 50-80% over the actual draw, so we recommend pairing nameplate with rack PDU readings if you have them. PDU output meters are the source of truth for actual instantaneous load.

Keep a column for each:

• Device (server, switch, storage, KVM)
• Quantity
• Nameplate watts
• Measured watts (from PDU)
• Power factor (typically 0.9 for modern IT, 0.7-0.8 for legacy)

For a modern server room with 30 servers averaging 200W each, plus 4 switches at 80W and 2 KVM at 30W, you have roughly 6,400W of measured IT load.

Step 2 — Add headroom

UPS systems should never run at 100% of nameplate. Beyond longevity issues, you lose redundancy headroom and the system has nowhere to go when a peak occurs. Industry rule of thumb is design for 70-80% load, leaving 20-30% margin.

Add a separate growth multiplier on top — typically 1.25× over a 3-year horizon, 1.5× over 5 years. Australian server rooms are growing in IT density faster than they shrink, even as virtualisation reduces server counts.

For our 6,400W example: 6,400 × 1.25 (3-year growth) ÷ 0.75 (75% target loading) = 10,667W of UPS capacity required.

Step 3 — Convert to kVA

UPS systems are rated in both kW and kVA. The relationship is kW = kVA × power factor. Modern UPS systems are typically 0.9-1.0 power factor at the output, but older units or sites with inductive loads are lower.

For planning, divide watts by 0.9 to get kVA. Our 10,667W example becomes 11.85 kVA. Round up to the nearest practical UPS size — typically a 12 kVA or 15 kVA single-phase unit.

Step 4 — Specify runtime

Runtime is the variable most clients underspec. The right runtime depends on what you're trying to accomplish:

• Generator handover — typical genset starts and stabilises in 10-15 seconds. Your UPS only needs ~30-60 seconds for safety margin if you have a generator.
• Graceful shutdown — if you're relying on UPS to shut down servers cleanly, allow 5-15 minutes depending on shutdown scripts.
• Ride through — if the site has a history of brief outages and no generator, 20-30 minutes is typical.
• Long runtime — telco exchanges and unmanned remote sites can need 4-8 hours.

Longer runtime requires either bigger battery cabinets or external battery extension modules. The cost curve gets steep above 30 minutes.

Step 5 — Choose topology

With size and runtime in hand, the last decision is topology:

• On-line double-conversion — best for sensitive equipment, dirty power, mission-critical loads. Always running through the inverter, so output is regenerated and isolated from input. Slightly less efficient (~92-96%) but the safest choice.
• Line-interactive — uses an AVR transformer for voltage regulation, switches to inverter only when input fails. Fine for typical IT, more efficient (~97-98%), substantially cheaper. Suitable when source power is reasonably clean.
• Standby/AVR — cheapest, lowest protection. Only suitable for non-critical loads. Not recommended for any production server room.

Three-phase considerations

Once you exceed about 20kVA, single-phase becomes impractical. The single-phase circuit current at 230V × 20kVA is around 87A — that's a substantial cable run and a heavy circuit breaker. Three-phase splits that across L1/L2/L3, dropping to about 29A per phase.

Three-phase UPS also balances load more evenly across the building electrical system, which the supply authority generally appreciates and which reduces phase imbalance issues for everything else on the switchboard.

A worked example

A mid-size SaaS company's server room in Sydney CBD: 24 servers averaging 280W (PDU measured) = 6,720W. Plus 6 switches at 90W, 2 SAN heads at 220W, KVM and misc 200W. Total measured load: 7,810W.

Apply 1.5× growth (5-year SaaS plan), 75% target loading: 7,810 × 1.5 ÷ 0.75 = 15,620W → 17.4 kVA at 0.9 PF. Round up to a 20 kVA online double-conversion UPS.

Runtime requirement: site has generator with 10-second start. Spec 30 minutes runtime to cover generator failure scenarios — that's an extended battery cabinet, but it's the difference between graceful shutdown and a hard crash if the genset doesn't start.

When to call us

UPS Services performs free load assessments for new server rooms and brownfield retrofits across Brisbane, Sydney and Melbourne. For larger projects we use IEC 61000-4-30 Class A meters to capture a full week of load data — measured beats estimated every time. [Request a Quote](/contact#quick-quote) and we'll come on site.

References

• AS/NZS 3000:2018 — Wiring Rules
• AS IEC 62040.3 — UPS performance classification
• IEEE 446 — Recommended practice for emergency and standby power systems