Data center backup power, explained
A data centre's promise is uptime, and uptime depends on what happens in the seconds after the grid fails. Backup power is the set of systems that carries the load through that gap and beyond it.
Why data centers need backup power
Most data centres draw their power from the public grid, and the grid can fail. Data Center Knowledge puts it plainly: backup power is a prerequisite for data center reliability and uptime. Downed lines, insufficient supply and summer brownouts all interrupt power.
The three layers of backup power
1. Uninterruptible power supply (UPS)
The UPS bridges the moment between grid loss and generator start. POWER magazine notes that it takes milliseconds for UPS to take over the supply of power to data center equipment, and most UPS systems allow only a few minutes, just enough for the backup generator to take over. See how this fits into the wider electrical chain in data center power distribution.
2. Standby generators
Standby generators are permanently installed and start automatically when grid power is lost. They are typically diesel reciprocating engines. POWER magazine reports that, depending on fuel availability, they can run for a day or up to 72 hours before refuelling is necessary.
Why diesel? Because it can be stored on site. Generators are sold with different ratings. A standby-rated set is built for occasional outage duty, not for running every day. The difference between standby, prime and continuous ratings is explained in prime vs standby power.
3. On-site power generation
The third layer is a power plant next to the load. Data Center Knowledge calls on-site generation the most reliable alternative to grid power for data centers, while noting that it is also the most expensive option. An on-site plant can serve as backup, or it can be the primary supply, with the grid as the backup. That second arrangement is what people mean by behind the meter power.
How Uptime Institute Tiers shape backup power
Uptime Institute's Tier classification sets the redundancy a facility needs. Even a Tier I site, the basic level, needs an uninterruptible power supply (UPS) for power sags, outages, and spikes and an engine generator for power outages. Tier II adds redundant capacity components. Tier III is concurrently maintainable. Tier IV adds fault tolerance.
The rules that limit backup generators in the US
Under the EPA's New Source Performance Standards for stationary compression-ignition engines (40 CFR 60.4211):
- There is no time limit on emergency use.
- Maintenance checks and readiness testing are capped at 100 hours per calendar year.
- Up to 50 of those hours may be non-emergency use, and those 50 hours generally cannot be used for peak shaving or to sell power.
From backup to prime power
A new pattern has emerged in AI data centres. POWER magazine reports that some developers are relying on backup generators for prime power until the grid interconnect becomes available. Running a generator as prime power is a different job from standby. It needs a prime- or continuous-rated machine, a non-emergency permit, a firm fuel supply and a maintenance regime built for thousands of hours a year. For the scale of load involved, see data center power requirements.
Where firm on-site power fits
If backup plant is going to run as primary power anyway, it is worth designing it as a power plant from day one. BridgePower Nuclear, a joint venture of USP&E and BAM Energy, is built around that idea.
First, USP&E supplies thermal power on gas, diesel, HFO or dual-fuel, grid-tied or behind the meter. It is backed by 24+ years in power, 150+ power projects in 45+ countries, 2,000+ MW in stock and 400+ engineers and project managers.
Read more in nuclear power for data centers.
If your site needs firm power before the grid can deliver it, talk to our team.
Frequently asked questions
What is the backup power for a data center?
Most data centres use layers. A UPS with batteries carries the load for seconds to minutes, standby generators start and carry it through longer outages, and some sites add an on-site power plant that can run the facility for extended periods.
How long can data center backup generators run?
As long as they have fuel. POWER magazine reports that diesel backup generators can typically run for a day, or up to 72 hours in some cases, before refuelling is needed. Sites on a gas pipeline can run longer.
Why do data centers use diesel generators for backup?
Diesel can be stored on site and is readily available, so the generator does not depend on a pipeline that could also be disrupted. The trade-off is air emissions and limits on non-emergency running hours.
How many hours a year can a backup generator run in the US?
Under the EPA rules for emergency stationary engines, there is no limit during emergencies. Maintenance, testing and certain non-emergency use are capped at 100 hours per calendar year, of which up to 50 hours may be non-emergency operation.
Can backup generators be used as primary power?
Some new AI data centres run generators as prime power while they wait for a grid connection. That changes the rating, permitting and fuel logistics, because standby-rated engines and emergency permits are not designed for continuous duty.