UPS in the data center: when you do NOT need one even if everyone says you do (and when a double-conversion unit costs you more than a stand-by)

There is a popular idea that says: your data center needs a UPS, and the better its topology (double conversion, for example), the better. Both halves of that statement are wrong, separately. There are loads that do not require a UPS, and others where a double-conversion unit charges you more in electricity than it saves you in incidents.

This article separates the two questions: when a UPS adds no value relative to the load, and when the difference between double conversion and stand-by is measured in kilowatt-hours per year, not in milliseconds of transfer time.

When you do NOT need a UPS (even when everyone says you do)

If the load tolerates a brief outage without data loss and without corruption, you do not need a UPS. We are talking about loads such as edge switches, label printers, WiFi access points, environmental monitoring displays, room lighting, or workstations running applications with local saving. What these loads require is general service continuity, not dedicated power quality.

There are also loads where the UPS is no longer the right solution. Compute equipment with distributed architectures that recover from another zone, file systems replicated in real time, ephemeral containers in clusters with active-active orchestration, or stateless services do not require a UPS per node: they require fault-tolerant design, and power is protected at another level (generator + transfer + power supply redundancy).

When you DO need a UPS (and what actually demands double conversion)

If the load demands zero transients at the output and active voltage regulation under any utility grid condition, then yes, you need a UPS. Medical equipment, SMT lines, instrumentation, servers in active-passive clusters with switchover time measured in microseconds, industrial process control equipment that responds to micro-variation. Stand-by does not apply here.

Loads with dual power supplies, or that can switch between two feeds in their own hardware, do not need double conversion in the UPS: their protection is local. What they do need is a UPS that delivers clean power during feed switchover, something a properly sized stand-by unit covers in most cases.

When double conversion costs you more than a stand-by

A double-conversion UPS works exactly as its name says: it converts AC to DC and then DC to AC continuously, delivering a reconstructed waveform at the output regardless of what happens at the input. Typical efficiency for models on the market runs from 92 to 95 %. A stand-by UPS (also called “line interactive” in its most complete version) only activates its inverter when the utility grid goes out of parameters, and its efficiency runs from 96 to 98 % in normal operation.

The difference adds up. For a 100 kW UPS operating 24×7, a 3 % efficiency gap means 7,300 kWh lost annually as heat above the stand-by unit’s minimum operating level. At an average medium-voltage DAC tariff in base hours, that extra kilowatt-hour is not a footnote: it is an annual cost on the order of a new pickup truck per year, for every 100 kW protected that does not require that power quality.

The calculation flips when the frequency of transients on the local grid increases. If the facility has variable frequency drives, furnaces, welding, or heavy inductive loads, transients on the utility line will be frequent and a double-conversion unit can degrade just the same depending on the filter and the robustness of the model. That is where the other decision appears: stand-by with an active regulation filter (line interactive with robust AVR) versus double conversion. In many cases, stand-by with AVR delivers equivalent quality at a lower operating cost.

How to decide the UPS type

Ask yourself three questions before choosing a topology:

1. Does the load tolerate a 4 to 10 ms transient without a problem, or does it need continuous regulation? If it tolerates it, stand-by or line interactive cover it. If not, double conversion.

2. How much does it cost to operate the annualized efficiency difference? If the figure exceeds the cost of the expected incidents, reconsider. If it covers it comfortably, double conversion is worth it.

3. What do you need during the transfer to the emergency generator? This is where double conversion wins most comparisons: the 10 to 30 seconds of switchover between the grid and the generator pass through a stand-by UPS without a support net, whereas a double-conversion unit absorbs them by delivering synthesized power with no variation.

What is worth sizing

For critical loads in production, the choice usually lands on double conversion: the transfer to the emergency generator is the scenario where the most failures are seen in operation. For auxiliary loads, environmental protection, engineering workstations, or fault-tolerant systems, a stand-by or line-interactive unit with batteries sized to the target runtime (no more) is the right choice and much cheaper to operate.

If you already have double conversion on loads where you do not need that power quality, you are paying extra consumption with no return in reliability. Migrating to a properly sized line-interactive unit recovers operating budget in less than a year of operation and frees electrical capacity for real growth.

Sources

[1] IEC 62040-3 — Uninterruptible power systems (UPS) — https://webstore.iec.ch/publication/5973

[2] Schneider Electric — EcoStruxure: UPS types and efficiency — https://www.se.com/ww/en/product-range/65746-galaxy-vl/

[3] Eaton — Double-conversion vs line-interactive UPS — https://www.eaton.com/us/en-us/products/backup-power-surge-it-power-distribution/backup-power-ups.html