Replace the battery or replace the UPS: a decision framework for Canadian IT and facilities buyers

Abstract and intended reader

When an aging uninterruptible power supply (UPS) starts alarming or fails a self-test, the practical question is rarely whether something must be spent. It is whether the money should buy a replacement battery string or a replacement unit. This white paper gives Canadian IT and facilities buyers a defensible way to answer that question: what the two aging clocks inside a UPS actually are, what evidence to collect before deciding, a break-even calculation you can run with your own quotes, and the field patterns a Canadian service organization applies when it faces the same choice.

The intended reader manages one UPS or a small fleet: an IT manager or system administrator with rack or tower units under 20 kVA, a facilities manager with a few three-phase machines, or a procurement lead comparing a battery quote against a unit quote. The framework applies to valve regulated lead-acid (VRLA) batteries in standby UPS duty. It is written from published manufacturer and standards evidence, with each material claim cited, and it states its limits: it is a decision framework, not a site-specific engineering study. Readers who only need the recommendation can jump to the decision table and the ten-step checklist near the end; the sections in between explain the evidence behind them.

The decision problem

Batteries are the component most likely to force this decision. A 2013 Ponemon Institute survey cited by Vertiv found that 55% of surveyed data centre operators named battery failure as a contributing cause of unplanned outages, and about one third of UPS system failures were linked to VRLA batteries; the survey allowed multiple causes, so these are contribution rates rather than shares of a single total [1]. VRLA batteries in UPS service typically need replacement every 3 to 5 years [1][3], while the UPS electronics around them are commonly planned around a 10 year service horizon for single-phase units [1]; larger three-phase platforms are often kept in service longer where parts and support remain available [9]. A UPS in service for 6 years, given the 3 to 5 year battery cycle, has typically already had at least one battery replacement and is often facing its next battery quote, and that quote is the moment the battery-versus-unit question gets asked.

The risks on the two sides of the decision are not symmetric. Put a new string into a chassis whose electronics fail a year later and most of the battery investment is stranded, because a string matched to one platform rarely transfers to its replacement. Replace a healthy unit because its batteries aged on schedule and you pay the full unit premium to fix a wear item that the manufacturer designed to be replaced. Both errors are common, and both are avoidable with the same small set of numbers: the battery quote, the unit quote, the credible remaining life of the existing electronics, and the expected life of a new string in the old chassis.

Terms, system boundary and assumptions

The framework uses these terms with fixed meanings. A UPS, also called an uninterruptible power supply, conditions power and carries the load from stored energy during an outage. The battery string is the series set of blocks or cells that provides that stored energy; its nominal DC voltage is set by the UPS bus. Runtime, or autonomy, is the battery-backed duration at a defined load. Design life is the laboratory classification of a battery family at a reference temperature; service life is what a specific installation actually achieves, and it is normally shorter. A UPS battery is commonly considered end of life at 80% of its rated capacity, and degradation typically accelerates beyond that point [1].

The system boundary is deliberate. The framework covers commercial single-phase UPS and small three-phase units with VRLA batteries in standby duty, whether internal cartridges or external strings and cabinets. It excludes flooded battery rooms, batteries used for cycling or grid services, generator integration questions, and any duty where the battery does more than ride through outages. Cartridge form factors and compatibility rules vary by platform, and the replacement battery compatibility and service-life help pages maintained by UPSPLUSBATTERY cover the common Canadian cases.

Three assumptions hold throughout. First, the load the UPS protects still matters to the organization, so doing nothing is not on the table. Second, quotes are available or obtainable for both paths. Third, the reader can establish the age of the unit and its batteries from records or nameplates; where that is impossible, the evidence section below treats the gap explicitly.

Why batteries and UPS electronics age differently

The decision is really a comparison of two remaining-life clocks, and they run at different speeds for different reasons.

The battery clock is chemical. VRLA batteries age through positive-grid corrosion, dry-out and rising internal resistance even when they sit fully charged and are almost never discharged. Published expectations for commodity VRLA in UPS float service centre on 3 to 5 years [1][3], and manufacturers classify commodity monobloc ranges at a 3 to 5 year design life, while long-life UPS ranges such as the EnerSys DataSafe HX class carry a 10 year design life at 25 C [6]. Temperature dominates the spread: Vertiv publishes that a controlled room near 25 C is needed to reach the 3 to 5 year lifespan, and that every 15 C of additional room temperature cuts the useful life of a typical VRLA battery in half [1]. A projection from that rule puts a battery specified for about 4 years at 25 C near 2.5 years at 35 C and near 2 years at 40 C. The projection is a rule of thumb, not a measured curve, and the exact battery datasheet governs; some manufacturers publish more conservative derating guidance.

Figure 1. Projected VRLA service life versus ambient temperature, derived from the published rule that life is halved for every 15 C above 25 C [1]. Calculated projection, not a measured curve; the battery datasheet governs.

The electronics clock is electrical and mechanical. DC bus and AC filter capacitors dry out and drift; a GDF synthesis of OEM preventive-maintenance service documentation places capacitor replacement planning in roughly the 12 to 15 year window and treats fans and air filters as scheduled wear items on shorter cycles [7]. Those figures are compiled from Vertiv and Eaton service documents, are model-specific, and were not independently re-verified at publication. A well maintained chassis can therefore outlive two or three battery strings, but it cannot do so indefinitely, and once a unit approaches its capacitor window the cost of keeping it credible rises sharply. Parts availability adds a hard edge to this clock: when a platform goes end of sale, replacement cartridges, fans and boards start aging out of distribution regardless of the condition of your specific unit.

Figure 2. Published lifecycle windows for UPS wear components [1][6][7]. Bars show typical planning ranges at reference conditions; the long-life VRLA point is the EnerSys DataSafe HX 10 year design life, and the capacitor window is a GDF synthesis of OEM service documentation that was not independently re-verified. Model-specific service documents govern.

Reading the two clocks together produces the core insight of this paper. A battery replacement buys you the shorter of two durations: the life of the new string, or the remaining life of the chassis it sits in. When the chassis has plenty of credible life left, battery money is well spent. When it does not, the same battery money buys very little protected time, and the unit path starts to win even though its sticker price is higher.

What evidence to collect before deciding

Battery age is a planning input, not a stand-alone failure diagnosis, and neither is a single alarm. A defensible decision integrates several evidence classes, in a sequence that matches IEEE recommended practice for VRLA maintenance and testing and the IEEE guide for UPS batteries [4][5][8].

Start with the records. Establish the unit's manufacture and installation dates, the date of the last battery replacement, the fault and alarm history, and whether the platform is still sold and supported. Then establish the load: measure or read the present load percentage, and note any growth planned for the next budget cycle. A UPS runtime calculator gives a quick first check of whether the runtime you observe is even plausible for the load you measured, before you attribute the shortfall to battery aging.

Then work up the evidence ladder. Visual inspection identifies swelling, leakage, corrosion or heat damage; any of these stops the analysis and triggers immediate qualified service. Operating voltage and ambient temperature establish charging context, and a room that has been running warm rewrites every life expectation from the previous section. Ohmic measurements (impedance, conductance or internal resistance) support trending against a baseline, and a unit that diverges from its peers is an investigation candidate; ohmic values do not convert directly into remaining runtime [8]. A UPS self-test provides model-specific functional evidence and can catch a disconnected or clearly weak string, but a short self-test pass does not prove contracted autonomy. Direct performance evidence comes only from a controlled discharge or runtime test to a defined endpoint, and that test needs a plan, an abort path and a recharge window.

Two cautions bound the testing itself. Repeated deep-discharge tests and runtime calibrations shorten battery life; manufacturer preventive-maintenance guidance compiled in the GDF synthesis warns against frequent battery calibration, so schedule discharge evidence deliberately rather than casually [7]. And a string-level pass does not clear a divergent block, an abnormal temperature or a charger defect; those findings survive a good runtime number and still need resolution [8].

The break-even calculation

The economics reduce to four numbers and one comparison. Define:

• B: installed cost of a full battery string replacement in the existing unit, parts plus labour, in CAD.
• U: installed cost of the replacement path, meaning a new UPS including its first battery set, in CAD.
• Lb: expected service life of the new string in the existing unit, in years. Use 3 to 5 for commodity VRLA at controlled temperature [1][3].
• R: credible remaining service life of the existing electronics, in years, from the evidence above.
• Lu: expected service life of the replacement unit, in years. Vertiv's single-phase comparison uses a 10 year horizon [1].

Estimate R from Figure 2 rather than guessing: take the electronics horizon for the unit type and subtract its current age, then reduce the result further if the platform is end of sale or the capacitor window is close. A single-phase unit planned around 10 years that is 6 years old starts near 4 years of nominal electronics life, less any deduction for end-of-sale parts risk or a hot room.

The battery path delivers protection for min(Lb, R) years, so its annualized cost is B divided by min(Lb, R). The replacement path costs U plus any battery replacements the new unit will itself need within Lu; for a VRLA unit on a 10 year horizon that is typically two more strings, while a lithium-ion unit typically reaches the same horizon on its original battery [1]. Divide that total by Lu for the annualized cost. The battery path wins while its annualized cost is lower, which happens when the existing electronics still have enough credible life:

R* = B x Lu / (U + future battery cost of the new unit)

R is the break-even remaining life. If your evidence supports more remaining electronics life than R, replace the batteries; if it does not, replace the unit. The formula is currency-neutral and works directly with your own Canadian quotes. For the B input, a companion article in this series describes what a UPS battery replacement costs in Canada and how the work proceeds, including calibration after the swap.

A worked example with published figures shows the shape of the result, though the figures come from a lithium-battery vendor's own promotional total-cost comparison and therefore favour the lithium case; use it only for the structure of the calculation, not as neutral pricing. Vertiv's 2023 single-phase comparison for a 1500 VA class unit lists a VRLA UPS at 800 USD, a lithium-ion UPS at 1,000 USD, and each VRLA string replacement at 450 USD including labour, with two replacements falling inside the 10 year horizon [1]. Against a new VRLA unit, the annualized replacement path is (800 + 900) / 10 = 170 USD per year, so R = 450 / 170, which is about 2.6 years. Against a new lithium-ion unit the annualized path is 100 USD per year and R = 4.5 years, which is beyond the 4 year life of the string in this example. Because the battery path's annualized cost cannot fall below B divided by Lb, at Lb = 4 it plateaus at about 113 USD per year, just above the lithium unit's 100, so a commodity 4 year string does not beat a new lithium unit on annualized cost. A string near the top of its range (Lb above about 4.5 years) in a chassis with at least that much remaining life can edge below it, so for a typical commodity string the decision against a new lithium unit usually turns on capital availability, compatibility and downtime rather than arithmetic [1].

Figure 3. Annualized cost of the battery path versus credible remaining electronics life, using Vertiv's published 2023 example figures (USD), which come from a lithium-vendor promotional comparison [1]. The battery path beats a new VRLA unit once about 2.6 years of electronics life remain; with a 4 year string it plateaus just above the lithium line, so beating a new lithium unit needs a longer-life string.

The calculation carries stated limits. It excludes downtime cost, disposal, freight and taxes; it assumes stranded battery value if the chassis fails first; and it is an ownership-cost comparison, not financial advice. Sensitivity is straightforward: R* scales linearly with B, so a battery quote at 30% of the unit quote pulls break-even down near 1.4 years, and one at 80% pushes it toward 3.8 years on the VRLA comparison. The unit's expected life matters too: if the replacement is credibly good for only 8 years rather than 10, both thresholds fall (to about 2.1 years against a VRLA unit and 3.6 years against a lithium unit), because the new unit's cost amortizes over fewer years.

Decision table: battery, battery upgrade, or new unit

Three paths cover nearly every real case, and the evidence plus the break-even number select among them.

Condition observed	Favoured path	Basis
Electronics with more than R* years of credible remaining life (for a 10 year single-phase unit, roughly its first 7 years, since 10 minus 2.6 is about 7.4), platform supported, batteries at or past 3 to 5 years	Replace the battery string in kind	Battery is the designed wear item [1][3]; annualized cost favours the string
Same as above, but runtime no longer meets the need	Battery upgrade within the same footprint, higher capacity blocks, after fit and charger verification	Field pattern: step capacity before enlarging the cabinet, once R exceeds R* [9]
Electronics near or past 10 years (single-phase), capacitor window approaching, or platform end of sale	Replace the unit; consider lithium-ion where the budget carries the premium	Electronics clock and parts availability dominate [1][7]
Repeated battery failures well before 3 years	Investigate temperature and charging first, then decide	Premature failure signals an environment or charger problem the next string will inherit [1][8]
Any swelling, leakage, heat damage or burning smell	Stop and call qualified service	Safety boundary, not an economic decision [8]

The battery path has one hard compliance edge. UL 1778, the North American safety standard for uninterruptible power systems, requires the UPS to identify the replacement battery manufacturer, catalogue number, quantity, string voltage and capacity, and UL states that replacing batteries with types not identified on the equipment and in the instructions is considered modifying the equipment [2]. When the framework lands on the battery path, exact-fit UPS replacement batteries matched to the manufacturer's identified type, quantity and string voltage keep the replacement inside the UL 1778 envelope. The same rule disposes of casual chemistry conversion: a lead-acid to lithium swap is acceptable only where the UPS manufacturer explicitly supports a lithium variant for that platform or the modified system is formally field-evaluated and re-listed [2].

Figure 4. The decision framework as a flowchart. The economic gate (R greater than R) sits ahead of the runtime split, so both battery outcomes pass it. Conceptual illustration of the process described in this paper; it does not replace the model-specific service manual.*

Field patterns from Canadian service work

The following patterns come from anonymized GDF Technologies service records covering roughly 450 UPS engineering responses across Canada in 2025 and 2026 [9]. They are recurring field decisions, not statistics, and they show the framework operating under real constraints.

Rule out the battery before condemning the electronics. When a very old UPS raised an error, the first diagnostic step was to ask when the batteries were last changed, because on an aged unit the fault is most often battery-driven. The inverse error is expensive: units get condemned on alarms that a string replacement would have cleared.

The string follows the bus voltage. Replacement strings are sized directly from the DC bus: a 192 VDC system takes 16 twelve-volt blocks, a 240 VDC system takes 20, and a 288 VDC system takes 24, all in series. Knowing the block count before requesting quotes makes the B input of the break-even calculation concrete.

Buy runtime inside the footprint first. Where a site wanted more autonomy from an existing installation, the string was moved to higher capacity blocks in the same cabinet after the cabinet manufacturer confirmed fit. Stepping capacity within the footprint defers both the cabinet build and the unit question, and it only proceeds after charger and weight verification.

A depleted-battery event is a decision trigger, not just an incident. On a monitored site, an outage that ran the battery to exhaustion prompted a proactive review: outage length, actual load, battery age and recharge behaviour. The review, not the event, decides whether the outcome was normal exhaustion beyond design autonomy or a battery at end of life.

Figure 5. Observed signals and the path each one favours. Summary of the evidence and decision sections; the safety-stop row is marked with a red triangle across all columns because it routes to qualified service rather than to any purchase path, and shapes distinguish the paths so the matrix reads without colour.

Failure modes, limitations and safety boundaries

Aged VRLA can fail without warning. Vertiv describes older batteries failing abruptly after appearing functional, which is why capacity at or below 80% is treated as end of life rather than as margin to consume [1]. Waiting for a visible failure is therefore not a strategy; it converts a planned replacement into an outage.

Partial fixes carry their own failure modes. Mixed-age strings and replacement of individual blocks change current sharing and charging behaviour, and they require UPS and battery OEM review rather than routine swapping [2][8]. Chemistry conversion without platform support is an equipment modification under UL's framing [2]. A second-hand or mismatched cartridge can pass a self-test and still fall short of the identified capacity.

The safety boundaries are firm. UPS battery strings store energy, and stored-energy, shock and arc hazards apply across the full range of UPS battery bus voltages, from small 24 to 96 VDC units to 192 VDC strings and higher; internal capacitors hold charge after disconnection. Physical signs such as swelling, leakage or heat damage end the do-it-yourself analysis immediately. Work inside three-phase units, external strings and battery cabinets belongs to qualified service personnel following the model-specific procedure, and spent lead-acid batteries must go to a proper recycling channel, which in Canada is well established for lead-acid chemistry.

The framework's own limits deserve the same clarity. It compares ownership cost and evidence-supported remaining life; it does not model downtime cost, which can dominate for revenue-critical loads. It assumes standby duty. It treats published lifecycle windows as planning ranges at reference conditions, not guarantees. And the worked example uses one manufacturer's 2023 USD figures, which will not match any specific Canadian quote in 2026.

Decision rule and checklist

The rule compresses to one sentence: replace the batteries when the evidence supports more remaining electronics life than the break-even value R*, and replace the unit when it does not or when a safety, compliance or parts-availability edge decides first.

The checklist operationalizes it:

1. Confirm there is no swelling, leakage, heat damage or burning smell; any of these means qualified service now.
2. Establish unit age, battery age, fault history and platform support status from records.
3. Measure the load and check observed runtime for plausibility against the load.
4. Verify the room temperature history; sustained heat shortens every battery estimate.
5. Obtain both quotes: string replacement (B) and unit replacement (U).
6. Estimate remaining electronics life (R) from the lifecycle windows minus current age, reduced for end-of-sale or a hot room.
7. Compute R* = B x Lu / (U plus the new unit's future battery cost) and compare with R.
8. On the battery path, order only the manufacturer-identified type, quantity and string voltage.
9. On the unit path, size for measured load plus confirmed growth, and weigh lithium-ion against VRLA on lifecycle cost.
10. Document the decision, the evidence and the date; the next owner of this question will thank you.

Single-phase owners who are still unsure after these steps can start from the battery help pages linked above before committing to either path. For three-phase systems, large strings or occupied critical sites, on-site UPS battery replacement service across Canada is available from GDF Technologies, the service organization behind this store. For everything else, the two quotes and this framework are usually enough to decide well.

Sources

1. Vertiv Group Corp. The Advantages of Using Lithium-Ion Batteries in Single-Phase UPS Applications for Edge Data Centers, SL-70595, rev 03/23. A manufacturer promotional total-cost-of-ownership comparison; its cost figures favour the lithium case and are used here only for calculation structure. https://www.vertiv.com/49ccd7/globalassets/shared/vertiv_liebert_psi5_lio_advantages-of-using-lithium-ions-batteries-sl-70595.pdf
2. UL Solutions. Uninterruptible Power Supply (UPS) System: Retrofitting/Replacing with New Batteries. https://www.ul.com/news/uninterruptible-power-supply-ups-system-retrofittingreplacing-new-batteries
3. Mitsubishi Electric Power Products, Inc. UPS Battery Maintenance: Monitoring, Service and Care. https://mitsubishicritical.com/critical-power-services/ups/battery-maintenance/
4. IEEE. IEEE 1188-2025, Recommended Practice for Maintenance, Testing, and Replacement of Valve-Regulated Lead-Acid (VRLA) Batteries for Stationary Applications. https://standards.ieee.org/ieee/1188/11656/
5. IEEE. IEEE 1184-2022, Guide for Batteries for Uninterruptible Power Supply Systems. https://standards.ieee.org/ieee/1184/6969/
6. EnerSys. DataSafe HX valve-regulated battery product page, which states a 10 year design life at 25 C. https://www.enersys.com/en/products/batteries/datasafe/datasafe-hx/
7. GDF Technologies internal knowledge library. Preventive Maintenance of UPS Systems (synthesis of Eaton, Vertiv, Schneider and Mitsubishi service documentation). The capacitor-window and calibration figures were compiled from those OEM documents and were not independently re-verified at publication, 2026.
8. GDF Technologies internal knowledge library. UPS Battery Testing and Condition Assessment (synthesis of IEEE 1188, IEEE 450, IEEE 1184 and OEM service documentation), 2026.
9. GDF Technologies. Internal service records, anonymized engineering and quoting patterns, 2025-2026.

Written on 11 July 2026. Prices and lifecycle figures cited from dated sources; verify against current quotes and datasheets before purchase decisions.