ANVILFIELD Try FieldOS

Datacenter

Data center battery and energy storage types field guide

The stored-energy choices behind critical power: VRLA, flooded, and lithium-ion (LFP vs NMC), plus the flywheel, supercapacitor, and BESS, and how footprint, life, cost, cooling, and fire risk drive the pick.

Battery Energy StorageLithium-IonVRLANFPA 855Data Center

Direct answer

A data center battery is the stored energy that carries the critical load from the moment utility power fails until the generator accepts the load, usually only minutes. The chemistry, VRLA lead-acid, lithium-ion, or a flywheel alternative, sets the footprint, life, cost, cooling, and fire risk. The manufacturer and applicable code control the specifics.

Key takeaways

  • A data center battery only bridges the load from utility failure to generator acceptance, commonly a few minutes up to around fifteen.
  • VRLA sealed lead-acid lasts about 3 to 5 years, while lithium-ion (LFP) often lasts 10 to 15 years in roughly a third to a fifth of the footprint.
  • LFP begins thermal runaway around 270 to 300 degrees C versus roughly 150 to 210 degrees C for NMC, making LFP the data center choice.
  • Lead-acid life roughly halves for every 10 degrees C of sustained operating temperature above the rated 25 degrees C (77 degrees F).
  • Lithium energy storage is listed to UL 9540, tested for fire propagation by UL 9540A, and governed by NFPA 855, commonly above a 20 kWh threshold.

What the battery does, and why the type matters

A data center battery is the stored energy that holds the critical load up through the gap between the utility failing and the generator picking it up. That is the whole job. It is not there to run the building. It is there to carry the load on the inverter for the few minutes it takes the standby generator to crank, build voltage, and accept the block, then hand off and recharge. How the UPS uses that energy, online versus line-interactive, is its own subject and lives in the UPS types guide. This guide is about what is in the cabinet behind it: the chemistry and the storage technology.

The type you pick is not a small decision dressed up as a big one. It drives the floor space the storage eats, the life before you replace it, the first cost and the long-run cost, the cooling the room needs, and the fire and life-safety case you have to satisfy. A lead-acid string and a lithium plant sized for the same minutes look nothing alike on the floor, on the maintenance schedule, or in the fire code.

Three families cover almost all of it. Lead-acid, in its sealed VRLA and vented flooded forms, was the data center default for decades. Lithium-ion, usually lithium iron phosphate in this service, is where new builds are going. And the no-battery options, the flywheel and the supercapacitor, trade chemistry for a spinning mass or a charged plate. Each carries the load a different length of time, and that is the first thing to get straight. Testing and maintaining whichever one you install is worked in the UPS battery maintenance and testing guide.

How long does the battery actually have to last?

Not long, and that surprises people. The data center battery only has to bridge the load from the instant of the power loss until the generator is online and carrying it, which is a matter of minutes, not hours. Design autonomy is usually short on purpose, commonly a few minutes up to around fifteen, sized to cover a generator start with margin for a start that does not go cleanly on the first crank. A generator that reaches load in seconds does not need an hour of battery behind it.

This short bridge is why the storage choice is about power for a short burst, not energy for a long haul. The battery has to deliver the full load current for those minutes, then it is done and the generator owns the outage. Size it to ride through to generator acceptance at the real design load, with headroom, because runtime falls off sharply as load rises and a number proven at light load is not the number you get at full load. The UPS guide covers the ride-through and the generator handoff from the UPS side.

Stretch the required runtime and the calculus changes. A site that wants longer autonomy, or that wants the battery to do more than ride-through, moves past the UPS battery into a larger battery energy storage system, which is a different animal with a different code basis. That distinction, power for the bridge versus energy for the long haul, runs through the rest of this guide.

VRLA: the sealed lead-acid default

VRLA, valve-regulated lead-acid, is the sealed lead-acid battery that was the data center standard for a generation and still fills a lot of rooms. Valve-regulated means the cell recombines its gas internally under normal charging and vents through a one-way relief valve only when it has to, so there is no watering and no open electrolyte. It comes in AGM, absorbed glass mat, where the electrolyte is held in a fiberglass separator, and in gel, where it is set in a silica gel, with AGM the common one in UPS service. It drops into a cabinet next to the UPS, holds a float charge, and asks for less hands-on care than a flooded string.

The trade-off is life and temperament. VRLA is commonly rated around 3 to 5 years of real service life in UPS duty, with premium pure-lead designs reaching closer to 10, but those are manufacturer figures that heat and discharge cycling pull down fast. It is heavy and it takes floor space, it is sensitive to temperature, and it can go into thermal runaway if it is overcharged or cooked. The failure is quiet: a VRLA string on float reads full voltage and gives no warning before it cannot make its minutes.

VRLA earns its place on first cost and simplicity. Where the budget is tight, the room is conditioned, and the operator is set up to test and replace strings on a schedule, it still works. The numbers, the float voltage, and the rated life come from the manufacturer's data, so size and set it to that, not to a rule of thumb.

Flooded (vented) lead-acid and the dedicated room

Flooded lead-acid, the vented or VLA cell, is the older wet-cell battery that buys long life with maintenance. The electrolyte is liquid, you can see the plates, and a single failed cell is replaceable, and a well-kept flooded string can run 15 to 20 years, longer than anything else on this list. That longevity is real, and it is why flooded cells still hold ground in some utility, telecom, and large institutional installations that were built around the maintenance program.

The cost of that life is upkeep and a room. Flooded cells gas hydrogen continuously on charge, so they lose water and need periodic watering, equalize charges to keep cells balanced, and specific-gravity readings to judge state of charge. The continuous hydrogen off-gassing is why they live in a dedicated, vented battery room with the full life-safety stack: forced ventilation to hold hydrogen below its flammable limit, eyewash and emergency shower for the sulfuric acid, spill containment, and neutralization on hand. The room and the ventilation are a topic in their own right, and the maintenance and the hydrogen handling are worked in the UPS battery maintenance and testing guide.

For a new data center, flooded has mostly faded. The footprint, the dedicated room, and the watering program cost more than the life saves against the alternatives, and the floor space is worth too much. Where it still appears, it is usually a legacy install or a site that already runs the maintenance discipline a flooded room demands.

Why lead-acid needs watching no matter the form

Both lead-acid forms share a failure pattern that decides how you live with them: they age quietly and they hate heat. Capacity drops cell by cell with no symptom on the front panel, and the recognized end of life is 80 percent of rated capacity, the point a string is replaced because capacity falls off quickly from there. Float voltage tells you almost nothing about that. A worn string reads full voltage right up to the outage it cannot carry.

That is why a lead-acid plant is a monitoring and replacement program, not a fit-and-forget install. You trend internal impedance to catch a weak cell early, you run a capacity discharge to prove the runtime is real, and you keep the room cool because heat is the single biggest thief of lead-acid life. The full discipline, the capacity test, the impedance trend, the float and temperature control, and the replacement criteria, is the subject of the UPS battery maintenance and testing guide. The point for choosing a type is that lead-acid carries an ongoing maintenance and replacement cost that the type comparison has to count, not just the first cost on the quote.

Why are data centers switching to lithium-ion?

Data centers are moving to lithium-ion because it holds far more energy in far less space and lasts longer, and the math on total cost has tipped in its favor. Lithium packs several times the energy of VRLA per pound and per cubic foot, so the battery room shrinks or the same room holds more autonomy. It lasts longer, often 10 to 15 years against the 3 to 5 typical of VRLA, recharges faster after a discharge, and tolerates a warmer room, which eases the cooling load on the space. Industry trackers now have lithium passing lead-acid as the dominant new UPS battery chemistry, and the push to dense AI halls is accelerating it.

The headline knock is first cost, and it is true: lithium costs more up front per kilowatt-hour than VRLA. The reason it still wins is total cost of ownership over the life of the install. A lithium plant that runs 10 to 15 years against lead-acid that needs replacing two or three times in that window, in less floor space, with less cooling and less maintenance, usually comes out ahead even though the purchase order is bigger. That is the modern trend in one line.

Lithium is not a free upgrade, though. It arrives with a battery management system that is part of the safety case, not an accessory, and it carries fire-detection and listing requirements that lead-acid does not. The density and the life are paid for with a more demanding safety and management layer, which the next sections work through. The room-scale and grid-scale version of lithium storage sits in the BESS work.

LFP vs NMC: which lithium chemistry?

The two lithium chemistries you will see are LFP, lithium iron phosphate, and NMC, nickel manganese cobalt, and for data center stationary storage LFP is the one that wins on safety. The split is between energy density and thermal stability. NMC holds more energy in less space and weight, which is why it dominates electric vehicles where every kilogram counts. LFP is a little less dense but far more thermally stable, and for a stationary battery sitting in a critical room, stability beats the last bit of density.

The stability gap is the heart of it. LFP's olivine crystal structure holds its oxygen in strong phosphorus-oxygen bonds that resist breaking down with heat, so an LFP cell begins thermal runaway at a much higher temperature, commonly cited around 270 to 300 degrees C, against roughly 150 to 210 degrees C for NMC. That wider margin means a smaller chance of a cell tipping into runaway, less violent behavior if one does, and a simpler, cheaper thermal and fire-protection design around it. LFP also tends to degrade more slowly, which adds to its life advantage. The exact onset temperatures vary by cell and manufacturer, so treat the figures as the published range, not a guarantee.

The practical read: NMC where energy per kilogram is the constraint, which is rarely the case for a fixed installation, and LFP where safety, cycle life, and a quieter thermal profile matter, which describes a data center. Most data center lithium UPS and storage products land on LFP for exactly that reason. Confirm the cell chemistry on the data sheet, because it changes the fire case and the listing behind the system.

The battery management system

Every lithium battery comes with a battery management system, the BMS, and it is part of the battery, not an add-on. Lithium cells have a narrow safe band for voltage, temperature, and current, and they do not tolerate being pushed outside it the way lead-acid shrugs off abuse. The BMS is the electronics that watch every cell or module, balance the charge across them so no cell runs ahead of the others, and trip the battery off line in milliseconds when a cell leaves its safe band. There is no equalize charge to run by hand and no per-cell float to set, because the BMS owns the cell-level management.

Because the BMS is the protection, commissioning a lithium plant is largely commissioning the BMS: confirming the cell telemetry is sane, that the addressing and communication are right, and that a forced limit actually opens the battery breaker rather than just lighting an alarm. The BMS also reports state of health, the cell's capacity now against new, which is the lithium version of the capacity number and replaces much of the discharge testing a lead-acid string needs.

One conflict to watch: the UPS charger and the BMS both think they own the battery, and their limits have to agree, so they get reconciled at commissioning. A BMS that disagrees with the charger, or whose own sensors were never verified, is a blind protection layer, which defeats the point of having one. The commissioning detail lives in the maintenance and testing guide.

What is thermal runaway, and how is lithium kept safe?

Thermal runaway is a self-feeding reaction where a lithium cell generates heat faster than it can shed it, the heat drives the cell hotter, and the reaction accelerates until the cell vents flammable gas and can ignite, with the heat able to propagate cell to cell through a module. It is the defining hazard of lithium storage and it is why lithium is its own fire-safety discipline rather than a drop-in for lead-acid. State it plainly: a lithium fire is a different and more serious event than a lead-acid one, and the design has to treat it that way.

The protection is layered and it is not optional. The BMS is the first line, catching the cell fault and opening the battery before a single cell runs away. Around that sits the fire-safety design the codes call for: off-gas detection that sniffs the hydrogen, carbon monoxide, and hydrocarbons a failing cell vents before it ignites, giving an early warning ahead of smoke or flame; spacing and clearance between units and from exposures so a fault in one does not cook the next; suppression and ventilation sized to the hazard; and deflagration venting where required. LFP's higher runaway onset, covered above, is part of why it is the chemistry of choice for this room.

The codes that govern it are specific. The system is generally listed to UL 9540, and the UL 9540A test method forces thermal runaway at the cell, module, unit, and installation levels to measure whether and how a fire propagates. That 9540A data is the evidence base the authority having jurisdiction uses to set spacing, vent sizing, and suppression under NFPA 855, the standard for stationary energy storage installation, which the International Fire Code references. NFPA 855 sets early-warning and off-gas expectations but does not prescribe specific detection setpoints, so those come from the equipment listing and the AHJ. Fire suppression for the room is a topic of its own; here the point is that the lithium fire case is designed in, not bolted on.

VRLA vs lithium-ion: which for the data center?

Lithium-ion wins the head-to-head for most new data center builds, and it wins on everything except first cost. The table below is the short version of the decision, and the reasoning under it is the same each row: lithium gives back floor space, years of life, and cooling, and asks for a higher purchase price and a more demanding fire case in return. VRLA still makes sense where first cost rules, the maintenance program exists, and the footprint is not precious.

The two rows that decide most projects are footprint and life. Lithium in roughly a half to a third of the installed space, lasting two to three times as long, is what flips the total-cost math even with the bigger up-front number. The row that gives engineers pause is safety: lithium's thermal-runaway hazard and its NFPA 855 obligations are real and have to be designed for, while lead-acid's hazard is hydrogen and acid, an old and well-understood problem. Neither is hazard-free. They are different hazards with different code stacks.

Read the table as a starting point, not a verdict for your project. The right answer depends on the runtime required, the floor space available, the cooling the room can give, the budget over the life of the install, and the fire-protection the building can support. Run those against the specific products, because the manufacturer's data is what the decision actually rides on.

AttributeVRLA (sealed lead-acid)Lithium-ion (LFP)
Footprint and weightBaseline; large and heavyAbout 1/3 to 1/5 of VRLA
Typical service lifeAbout 3 to 5 years, up to ~10 premiumOften 10 to 15 years
First costLowerHigher
Total cost over lifeHigher; replaced 2 to 3 timesOften lower despite first cost
CoolingTight; heat halves lifeTolerates a warmer room
CyclesLimited; cycling shortens lifeMany more deep cycles
Key safety caseHydrogen, acid, thermal runaway on overchargeCell thermal runaway, BMS, UL 9540 / NFPA 855

The flywheel: ride-through with no chemistry

A flywheel stores energy in a spinning mass instead of a chemical cell, and it rides through an outage on inertia. A motor keeps a heavy rotor turning at high speed, and when the utility drops, the rotor's momentum drives a generator that carries the load for the seconds it takes the standby generator to start and accept it. The energy reserve is small compared to a battery bank, so ride-through is short by design, commonly on the order of 10 to 30 seconds at full load. That is enough, because most utility disturbances last only seconds and a well-maintained generator starts inside that window.

The pull of the flywheel is everything a battery is not. No cells to age, no electrolyte, no battery room, no thermal-runaway risk, and a known, repeatable ride-through that does not fade with calendar life the way a battery does. It tolerates a wide temperature range and cycles endlessly without wear from cycling. On sites that also keep batteries, a flywheel is sometimes set in front as the first defense against short transients, absorbing the brief dips so the batteries cycle far less and last longer.

The cost is that the energy reserve is measured in seconds. A flywheel design leans hard on the generator starting reliably and fast, so it pairs a thin ride-through with a tight generator start spec, and some sites back it with a short battery for margin. The rotary idea taken further, the diesel rotary UPS that couples the spinning mass to a diesel engine on one shaft, is worked from the UPS side in the UPS types guide.

Supercapacitors and very short ride-through

Supercapacitors, also called ultracapacitors, store energy electrostatically on charged plates rather than in a chemical reaction, and they sit at the far short end of the runtime scale. They charge and discharge almost instantly, deliver very high power for a very brief moment, and survive enormous numbers of cycles with little wear, far more than any battery. The catch is energy: a supercapacitor holds little of it, so ride-through is seconds at most, shorter than a flywheel and far shorter than a battery.

That makes them a niche tool in critical power rather than a main energy store. They suit riding through the briefest disturbances, the sub-second sags and transients, and they pair well with a battery or a generator that handles the longer events, carrying the first instant while the slower source comes up. Their high cycle life means they shrug off the constant small dips that would wear a battery. Where a site needs power for the blink of an outage and energy for the rest, a supercapacitor on the front with a battery or generator behind it is the pattern. As a standalone bridge to a generator, they rarely carry enough energy.

What is a BESS, and how is it different from the UPS battery?

A BESS, a battery energy storage system, is a larger, grid-interactive battery that does more than ride through to the generator, and it is the storage trend reshaping data center power. Where the UPS battery is built for power, a short, hard burst of minutes, the BESS is built for energy, sustained output over a much longer window. It is sized in megawatt-hours, it talks to the grid, and it does jobs the UPS battery never touches: peak shaving to cut demand charges, often cited at 20 to 40 percent off peak load, load shifting to buy energy cheap and use it dear, frequency regulation and other grid services, spinning reserve, and even black start.

The driver right now is the grid connection itself. Data centers, especially the dense AI campuses, are waiting in long utility interconnection queues, and an on-site BESS lets a site cap its grid draw, flatten the spikes from AI training loads, and in some cases get connected and operating faster than the full utility upgrade would allow. The BESS becomes a tool for getting power to the building, not just backing it up. That is a different value case than ride-through, and it is why BESS is showing up alongside, not instead of, the UPS.

The two are increasingly run together as layers: the UPS for the instant, milliseconds-fast, never-a-gap protection of the IT load, and the BESS for the longer-duration energy and the grid economics behind it. A BESS is almost always lithium, usually LFP, and it carries the full UL 9540 and NFPA 855 fire and installation stack because of its size. The detailed commissioning of a room-scale or grid-scale BESS is its own discipline.

Power for the bridge versus energy for the long haul

The cleanest way to keep the storage types straight is to ask whether the application needs power or energy. Power is a hard burst delivered fast for a short time. Energy is a steady output sustained over a long time. The UPS battery, the flywheel, and the supercapacitor are power devices, sized to carry the load for seconds to minutes until the generator takes over. The BESS is an energy device, sized to run for hours and to play in the energy market.

Mixing the two up is how a storage system ends up wrong for its job. A flywheel asked to carry a long outage will be empty before the generator is late. A BESS asked to be the instant ride-through is slower and more complex than the job needs. Size and choose against the question the storage actually answers. The ride-through to the generator is a power problem with a short, witnessed runtime. The demand-charge and grid-services case is an energy problem with a duration measured in hours. The hybrid sites run both because they have both problems.

Cabinet batteries versus a dedicated battery room

Storage gets deployed two ways, and the choice follows the size and the chemistry. Small and mid-size installs put the battery in a cabinet, often integrated with or sitting right beside the UPS, in racks or matched enclosures on the white space or in the electrical room. Cabinet batteries are quick to install, easy to expand by adding cabinets, and keep the storage near the load it protects. Most lithium UPS plants and many VRLA ones live this way.

Large installs, and almost all flooded ones, go in a dedicated battery room. A room gives the floor loading for heavy lead-acid, the ventilation a chemistry needs, and the separation the fire code wants for a big lithium plant, and it keeps the maintenance and the hazard out of the white space. Flooded lead-acid effectively requires its own room for the continuous hydrogen off-gassing and the acid handling. A large lithium plant or a BESS is room or container scale and brings its own NFPA 855 spacing, detection, and suppression with it.

The deployment is not just a packaging choice. A cabinet next to a UPS and a dedicated battery room have different cooling, different fire protection, and different floor loading, and the storage type tends to force the answer. Heavy or gassing chemistries push toward a room. Compact lithium can sit in a cabinet until the plant gets big enough that the fire case sends it to a room of its own.

Charging, float, and the recharge after a discharge

How a storage type is charged is part of the type, and it differs sharply across chemistries. Lead-acid sits on a float charge, a steady voltage that holds it full without overcharging, commonly around 2.25 to 2.30 volts per cell for VRLA at the reference temperature and set to the manufacturer's curve. That float has to be temperature-compensated, lowered as the battery warms, because a fixed float on a hot string is the classic path into thermal runaway. Flooded cells also take a periodic equalize or boost charge to balance the cells. Lithium does not float in that sense at all: the charger runs a constant-current then constant-voltage profile and backs off, and the BMS balances the cells. The float and temperature-compensation detail is worked in the maintenance and testing guide.

Recharge after a discharge is the part that bites the unwary. When the battery has carried an outage and the generator lands, the charger goes to work refilling it, and that recharge is a real load on top of the load the generator already picked up. It is a step the generator has to swallow, and lead-acid is slow to recharge while lithium is fast, which changes both the generator sizing and how long the storage is unprotected before it is ready for the next event. Size the source for the recharge step, often with a walk-in or soft-start charge so the generator is not slammed, and remember the storage is not back in service until it is recharged.

How does temperature affect the storage choice?

Heat shortens lead-acid life sharply, and that single fact shapes both the room and the type decision. For lead-acid, the rule of thumb is brutal: service life roughly halves for every 10 degrees C, about 15 to 18 degrees F, of sustained operating temperature above the rated 25 degrees C, or 77 degrees F. A VRLA string rated for 5 years at 77 degrees F can be down to half that if it lives at 95 degrees F. That makes battery room cooling a design item, not a comfort setting, and it adds the cooling load to the cost of a lead-acid plant.

Lithium changes the equation by tolerating a warmer ambient than lead-acid, which lets the room run warmer and cuts the cooling the storage demands. It is not immune, its own life and warranty track temperature too, but the headroom is wider, and for a hot or cooling-constrained site that headroom can decide the type. The temperature side of testing and trending, and the Arrhenius mechanism behind the lead-acid number, are covered in the UPS battery maintenance and testing guide.

The selection lesson is to count the cooling as part of the storage decision. A lead-acid plant that needs the room held cool to make its life carries a cooling cost a lithium plant does not, and a room that drifts warm replaces lead-acid strings years early. Build the cooling, or pick the chemistry that needs less of it.

Monitoring the storage

Whatever the type, the storage has to be watched, because the failure mode of stored energy is silence. A battery monitoring system trends cell or jar voltage, internal impedance, temperature, and string current continuously and alarms on a problem instead of waiting for the next manual check. For lead-acid, the impedance trend is the early warning that catches a weak cell before a capacity test would, and the temperature and current trends are what flag a string drifting toward thermal runaway. For lithium, the BMS is the monitor, reporting cell telemetry and state of health and tripping on a fault.

The value is in the trend and the alarm, not the dashboard, and the monitor's own sensors get verified at commissioning so you are not trusting a blind gauge. The detail of impedance baselines, capacity trending, and what the numbers mean lives in the UPS battery maintenance and testing guide. For the type decision, the point is that every storage type needs a way to see the slow, quiet degradation, and lithium bundles that watching into the BMS while lead-acid needs a monitoring system added around it.

End of life, replacement, and recycling

The battery is a consumable, and the replacement cycle is part of what you are buying. Lead-acid reaches end of life at 80 percent of rated capacity and gets replaced as a whole string, not chased cell by cell, because a fresh jar in an old string makes a mismatch that works the old cells harder. That replacement comes around every few years for VRLA, which is the recurring cost the type comparison has to count. Lithium runs far longer, 10 to 15 years, but it too reaches an end of life when state of health falls past the spec or warranty threshold, and the whole plant gets planned for replacement.

Disposal is regulated and the two chemistries are handled differently. Lead-acid is one of the most recycled products there is, with the lead and the case going back through the manufacturer or a certified recycler, never the dumpster. Lithium has its own transport, handling, and recycling rules, and a damaged or swollen lithium cell is a hazard in transit that follows a specific procedure. Document the removal and the disposal chain of custody for both, because the record matters for the environmental rules and the warranty.

Plan the replacement before the failure, not after it. Order the matched replacement against the capacity or state-of-health trend, and schedule the swap with the UPS on bypass or with redundant capacity carrying the load so the room is never unprotected during the change. A storage type with a short life is not disqualified by it, but the replacement cadence and the disposal handling belong in the cost and the plan from the start.

The codes that govern the battery

Which codes apply depends on the chemistry, and getting that wrong is a real exposure. Lead-acid installations fall under the NEC storage-battery provisions in Article 480, while a lithium BESS is governed by Article 706 for energy storage systems, the fire code, and the ventilation practice for hydrogen, with IEEE recommended practices framing the installation and the room. Flooded rooms carry the heavier life-safety stack for the liquid electrolyte and the continuous hydrogen.

Lithium brings the energy-storage fire and life-safety standards. The system is generally listed to UL 9540, the separate UL 9540A test method evaluates thermal-runaway fire propagation, and NFPA 855, the standard for stationary energy storage installation, governs spacing, detection, ventilation, and suppression where it applies, commonly above a 20 kWh threshold, with the International Fire Code referencing it. NFPA 855 has been through edition updates, including a 2026 edition, so the obligations move over time and the adopted edition controls.

Above all of these sit the manufacturer's instructions and the project specification, which set the actual numbers, and the authority having jurisdiction, which has the final say on what is enforced for this building. Confirm the section and the edition against the adopted code before citing them on a submittal, and where a standard and the spec disagree, the stricter controlling document wins. Treat the code basis as part of choosing the type, because a chemistry that triggers NFPA 855 changes the room, the budget, and the schedule.

Selecting the storage for the data center

Pick the storage by running five questions against the project and the specific products, in roughly this order: the runtime required, the footprint available, the life and total cost over the horizon, the cooling the room can give, and the fire and life-safety the building can support. The runtime answers the first fork. A short bridge to the generator is a power problem, so a UPS battery or a flywheel. A long autonomy or a grid-services role is an energy problem, so a BESS. Most data center ride-through lands on a battery, and for a new build that battery is increasingly lithium iron phosphate.

From there the type follows the constraints. Tight floor space, a long life horizon, and a warm or cooling-constrained room push to lithium LFP, accepting the higher first cost for the lower total cost, and budgeting the NFPA 855 fire case. A tight first-cost budget, an existing maintenance program, and floor space to spare can still justify VRLA. A site that wants no battery room and a known, repeatable ride-through, leaning on a fast generator, looks at the flywheel. A flooded plant is a legacy or a long-life institutional call, not a default for a new hall.

The decision is a tradeoff, not a winner, and the manufacturer's data is what it rides on. Run the runtime at the real design load, count the replacement cycles and the cooling in the total cost, and confirm the fire-protection the chemistry forces against what the building can give. Then document why, because the next engineer needs to see the basis, not guess at it.

AI power density and the storage it pushes

The jump to AI and high-density compute is changing the storage picture faster than anything in years. Rack power that used to be a handful of kilowatts is now tens, and a hall full of training accelerators draws hard, swings its load fast, and strains both the room and the utility feed. That density rewards storage that is compact, that handles deep and frequent cycling, and that can help manage the load the building presents to the grid, which is exactly the profile of lithium iron phosphate.

It is also pulling the BESS into mainstream data center design. The AI campuses sitting in long utility interconnection queues are using on-site batteries to cap and flatten their grid draw, ride the spikes from training jobs, and in some cases get connected faster than a full grid upgrade would allow. The result is a layered architecture: lithium UPS storage for the instant ride-through of the IT load, and a larger BESS for the energy and the grid economics behind it. The direction of travel is dense, lithium, and increasingly grid-aware, and the storage decision for a new build is being made against that, not against the data hall of ten years ago.

What to document

The storage selection is defensible only if the record says what was chosen and why, so the next engineer and the operations team can see the basis instead of guessing at it. A one-line that shows a battery but never states its type, chemistry, runtime basis, and the code it was designed to leaves everyone in the dark, and the dark is where a replacement gets ordered wrong or a fire case gets missed.

Capture the storage type and chemistry, the runtime proven at the real design load, the footprint and floor loading, the rated and expected life with the replacement plan, the cooling the room provides, the safety and fire basis including the listing and the code that applies, and the monitoring or BMS in place. Tie each entry to the basis of design so a later change gets checked against the intent. The table is the short list of what the record needs.

Field to recordWhy it matters
Storage type and chemistrySets the maintenance, the life, and the code stack
Runtime proven at design loadThe bridge to the generator that was actually bought
Footprint and floor loadingWhat the storage costs in space and structure
Rated and expected life, replacement planThe recurring cost and the cadence
Cooling and room temperatureDrives lead-acid life and the cooling cost
Safety and fire basis (UL 9540, NFPA 855)The listing and code the install satisfies
Monitoring system or BMSHow the quiet degradation gets seen

Common mistakes

  • Putting lead-acid in a hot room and letting heat halve the life without anyone tracking why strings die early.
  • Installing lithium with no thermal-runaway plan, no off-gas detection, and no BMS verification at commissioning.
  • Ignoring NFPA 855, UL 9540, and the AHJ for a lithium plant and discovering the fire case at inspection.
  • Sizing the runtime short of the real worst-case generator start and load-acceptance time at full load.
  • Running no monitoring, so a weak string is found at the outage instead of on the impedance or state-of-health trend.
  • Dropping a fresh jar into an aged lead-acid string and creating a mismatch instead of replacing the string.
  • Counting only first cost and missing that VRLA replaced two or three times costs more than lithium over the life.
  • Treating a BESS as the instant ride-through, or a flywheel as a long-autonomy store, instead of matching power versus energy to the job.
  • Forgetting the recharge step is a real load on the generator and leaving the storage unprotected while it refills.
  • Having no end-of-life or recycling plan, so replacement and the regulated disposal become a scramble.

Field checklist

0 of 9 complete

Want this checklist to run itself on every job — with photo proof and a signed record crews can hand the customer? That's FieldOS.

Standards and references

The standards split by chemistry, and naming the right one for the point is what separates a credible basis from a guess. For lead-acid, the IEEE recommended practices frame the installation and the room: IEEE 1188 and IEEE 1187 for valve-regulated lead-acid maintenance and installation, IEEE 450 and IEEE 484 for vented (flooded) cells, and IEEE 485 and IEEE 1184 for sizing. IEEE 1635 with ASHRAE Guideline 21 covers ventilation and thermal management of stationary battery installations. The testing and maintenance side of these is worked in the UPS battery maintenance and testing guide.

Lithium and energy storage carry the fire and life-safety stack. The system is generally listed to UL 9540, the UL 9540A test method evaluates thermal-runaway fire propagation, UL 1973 covers batteries for stationary and similar applications, and NFPA 855 governs the installation of stationary energy storage where it applies, with the International Fire Code referencing it. NFPA 855 has moved through editions, including a 2026 edition, so confirm the adopted one. The NEC, NFPA 70, covers the electrical installation, with Article 480 covering storage batteries and Article 706 governing stationary energy storage systems (the governing article for a lithium BESS), and NFPA 70E covers the arc-flash and shock work practices on the DC string.

Above all of these sit the manufacturer's documentation and the project basis of design, which set the actual life, voltages, spacing, and runtime, and the authority having jurisdiction, which has the final say on what is enforced for this building. Hedge the life, the float and charge voltages, and the spacing to the manufacturer and the adopted code, not to a rule of thumb, and confirm every section and edition before citing it on a submittal. Where a standard and the spec disagree, the stricter controlling document wins.

Units and terms

Storage work runs on a few units and a lot of overlapping names, and reading the wrong one is how the wrong type gets sized. Capacity is in ampere-hours or kilowatt-hours, and runtime is the minutes the store carries the load at a given discharge rate. Power is the rate of delivery in kilowatts or megawatts; energy is the total delivered over time in kilowatt-hours or megawatt-hours. The difference between a power device and an energy device is the difference between the UPS battery and the BESS.

Keep the chemistry and technology names straight, because they decide the standard and the procedure. VRLA is sealed valve-regulated lead-acid, in AGM or gel form. VLA is vented flooded lead-acid. Lithium-ion in this service is usually LFP, lithium iron phosphate, chosen over NMC, nickel manganese cobalt, for thermal stability. A flywheel stores energy in a spinning mass, a supercapacitor stores it electrostatically, and a BESS is a large grid-interactive battery system. Thermal runaway is the self-feeding heat reaction that destroys a cell, and state of health is the lithium measure of capacity now against new.

VRLA / VLA
Valve-regulated (sealed) lead-acid and vented (flooded) lead-acid, the two lead-acid families with different maintenance and rooms
Lithium-ion (LFP / NMC)
Lithium iron phosphate, favored in data centers for thermal stability, versus the denser but less stable nickel manganese cobalt
BMS
Battery management system, the electronics that monitor, balance, and protect lithium cells and report state of health
Thermal runaway
A self-feeding reaction where a cell heats faster than it can cool until it vents flammable gas and can ignite
Flywheel
Kinetic storage that rides through on a spinning mass, commonly 10 to 30 seconds, with no chemistry
Supercapacitor
Electrostatic storage delivering very high power for seconds with very long cycle life and little energy
BESS
Battery energy storage system, a large grid-interactive battery for energy, peak shaving, and grid services
Power vs energy
Power is a short hard burst (ride-through); energy is sustained output over hours (autonomy and grid services)

Related tools

Calculators and readiness checks for this work

Compare your options

FAQ

What kind of batteries do data centers use?

Data centers use VRLA sealed lead-acid, the long-time default, and increasingly lithium-ion, usually lithium iron phosphate, for its smaller footprint and longer life. Flooded lead-acid appears mainly in legacy or long-life installs. Flywheels and supercapacitors handle short ride-through without chemistry. The battery only has to bridge minutes to the generator.

What is the difference between VRLA and lithium-ion batteries?

VRLA is sealed lead-acid: lower first cost, but heavy, large, and lasting only about 3 to 5 years. Lithium-ion holds three to five times the energy in the same space, lasts 10 to 15 years, tolerates a warmer room, and needs a battery management system and an NFPA 855 fire case. Lithium usually wins on total cost.

Why are data centers switching to lithium-ion?

Data centers are switching to lithium-ion because it packs far more energy in less space and weight, lasts two to three times longer than VRLA, recharges faster, and tolerates a warmer room, which lowers total cost despite a higher first cost. The push to dense AI halls, where space and cycling matter, is accelerating the shift to lithium iron phosphate.

What is thermal runaway in a lithium battery?

Thermal runaway is a self-feeding reaction where a lithium cell generates heat faster than it can shed it, driving itself hotter until it vents flammable gas and can ignite, with the fire able to spread cell to cell. It is the defining lithium hazard, managed by the BMS, off-gas detection, spacing, and suppression under UL 9540 and NFPA 855.

LFP or NMC: which lithium chemistry for a data center?

LFP, lithium iron phosphate, is preferred for data centers because it is far more thermally stable than NMC, beginning thermal runaway around 270 to 300 degrees C versus roughly 150 to 210 degrees C for NMC. That wider safety margin simplifies the fire design. NMC is denser, which matters for vehicles, not for a fixed installation where safety wins.

How long does a data center UPS battery need to last on an outage?

Only minutes. The data center battery has to carry the load from the moment utility power fails until the generator starts and accepts it, commonly a few minutes up to around fifteen. It is sized for a short power burst, not long autonomy. Prove the runtime at the real design load, because runtime falls off sharply as load rises.

What is the difference between a UPS battery and a BESS?

A UPS battery is a power device that carries the load for minutes until the generator takes over. A BESS, battery energy storage system, is an energy device that runs for hours and does peak shaving and grid services. Data centers increasingly run both as layers: the UPS for the instant, the BESS for the endurance.

What is a flywheel UPS and how long does it last?

A flywheel UPS stores energy in a spinning mass and rides through an outage on inertia, commonly 10 to 30 seconds, long enough for a generator to start. The appeal is no chemistry: no cells to age, no battery room, and no thermal-runaway risk. The cost is a short reserve that leans hard on a fast, reliable generator start.

Do data center batteries need a dedicated room?

It depends on the type and size. Compact lithium and many VRLA plants sit in cabinets beside the UPS. Flooded lead-acid needs a dedicated, vented room for continuous hydrogen off-gassing and acid handling. A large lithium plant or a BESS moves to a room scale for NFPA 855 spacing, detection, and suppression. The chemistry usually forces the answer.

What codes apply to lithium-ion energy storage in data centers?

Lithium energy storage is generally listed to UL 9540, with the UL 9540A test method evaluating thermal-runaway fire propagation, and NFPA 855 governs the installation, spacing, detection, and suppression where it applies, commonly above a 20 kWh threshold, with the International Fire Code referencing it. UL 1973 and NEC Article 480 also apply. The adopted edition and the AHJ control.

People also ask

Codes cited in this guide

This guide is written and reviewed against the published standards below. Always confirm the current adopted edition with the authority having jurisdiction.