Battery Storage in Data Centres: Where It Wins, Fights, and Loses

The Volta Foundation’s latest report on battery energy storage in data centres does not make the argument most headlines assume. It does not claim that batteries have won. It claims that batteries are winning in some applications, fighting for ground in others, and still losing in a few, and that treating those three conditions as one is the most expensive error an energy investor or infrastructure developer can make right now.

The context behind that claim is a rare convergence of pressures. AI infrastructure build-out has produced interconnection queues stretching years, gas turbine deliveries pushed into the next decade, and hyperscalers with the capital and urgency to build their own power assets. Battery storage sits directly beneath all three.

But the competitive picture is segmented by application, duration, and geography in ways that aggregate market forecasts simply flatten out.

Here is what this analysis maps: which segments of the data centre power market are genuinely opening up for batteries, which remain contested, and which still belong to diesel and gas. Read it as a way to calibrate your exposure to battery supply chains, grid infrastructure plays, and energy transition positions with the precision the segmentation demands.

Where battery storage already has the upper hand

Start with the wins, because they are real and specific. Battery energy storage systems (BESS) outperform gas turbines and diesel in a defined set of applications, and the reason is consistent across all of them: speed and daily cycling frequency, not a simple cost advantage per unit of energy.

The four application areas where batteries compete strongly, each for a distinct reason, are these:

• Curtailment flexibility: batteries actively balance loads against constrained grid connections, allowing a facility to keep operating when the grid cannot deliver full power.
• UPS ride-through: immediate backup at the moment of a grid fault, where sub-second response is the entire requirement.
• Grid services: exporting power or providing demand response back to the regional grid, turning a cost centre into a revenue line.
• Sub-second power quality: frequency and voltage stabilisation that fossil alternatives physically cannot match on response time.

Each of those wins is conditional on its application context. “BESS is competitive” only means something at the segment level, which is precisely why aggregate market figures mislead.

Market sizing context (unverified analyst projection) MarketsandMarkets values the dedicated data centre BESS market at approximately US$19.4 billion in 2026, reaching US$34.8 billion by 2031, a 19.5% CAGR. Treat this as an analyst estimate, not independently confirmed.

The commercially significant win right now is bridge power. FERC Order 2023’s interconnection reform has already triggered the withdrawal of over 1,800 GW of non-viable projects (unverified), and the queue backlog means new fossil generation cannot connect fast enough to serve AI load on the timeline hyperscalers need.

Batteries offer a time-to-power advantage over gas and diesel for any asset stalled in that queue. That is not a policy abstraction. It is a near-term procurement window with direct revenue implications for battery supply chain investors, because these four categories are where purchase orders are being written, not where feasibility studies are being run.

The interconnection queue dynamics driving this procurement window are not evenly distributed across grid regions; FERC Order 2023’s restructuring accelerated withdrawals in congested markets like PJM and MISO while leaving some regional queues largely intact, which means the time-to-power advantage batteries hold varies materially by geography.

Understanding the competitive map: where batteries win, where they fight, and where they lose

To read the competitive picture correctly, drop the binary of batteries versus diesel and think instead in terms of duration and load profile. The technology verdict is not fixed. It changes with how long you need backup to last.

Below roughly eight hours of backup duration, BESS economics generally beat diesel once fuel and maintenance costs are fully counted, according to estimates flagged as unverified. Beyond that point, the outcome becomes location-specific rather than settled either way.

The capital cost comparison shows why the debate stays live. National Renewable Energy Laboratory estimates place diesel generators at around US$1,000 per kW against roughly US$1,300 per kW for a four-hour battery system (both unverified).

Diesel’s residual advantage is operational: a rapid start of 10-15 seconds to full load (unverified), high energy density, and indefinite refuelling. Those attributes keep it relevant for the multi-day events that batteries cannot economically cover.

The hybrid architecture consensus

The resolution the industry has landed on is not either-or. It is both.

Medium-voltage UPS architecture has become the preferred integration path for hyperscalers pursuing the hybrid model, because it allows a single battery system to serve both the millisecond ride-through function and the daily cycling role without requiring separate inverter infrastructure for each application.

The emerging standard pairs a 1-4 hour battery for daily cycling and standard outages with a smaller, right-sized diesel or gas unit reserved strictly for rare multi-day continuity events. The logic sits on a single observation: the overwhelming majority of grid disturbances are short, so a 1-4 hour battery covers most outage scenarios while earning revenue the rest of the time, whereas diesel sits idle as a cost.

For an investor sizing a battery company’s data centre exposure, that hybrid consensus clarifies rather than expands the opportunity. The addressable market is not the entire backup power budget. It is the sub-eight-hour backup and active grid services segment, which is large but materially narrower than the headline figures most market research quotes.

One further caveat matters. The economics of peak shaving and grid services depend heavily on regional demand-charge rate structures. In markets without them, a key financial benefit attributed to BESS simply disappears, which is why the same technology looks compelling in one jurisdiction and marginal in another.

What hyperscaler procurement reveals about the trajectory

Individual battery announcements read as press releases. Taken together, the hyperscaler deployments form a pattern, and the pattern is the signal. These operators have the technical expertise and the capital to make architecture decisions that smaller operators will follow, which makes their procurement a leading indicator of where the competitive balance is heading.

The deployment evidence, all flagged as unverified, breaks down by operator:

• Google: a 300 MW / 30 GWh Form Energy iron-air system with Xcel Energy in Minnesota; a 2.75 MW / 5.5 MWh Fluence battery in Belgium cutting diesel use and trading demand response; more than 100 million lithium-ion cells across its global fleet.
• Microsoft: replaced diesel backup at a Swedish data centre with a 16 MWh Saft BESS; runs a grid-interactive Eaton UPS and lithium-ion model in Dublin that it intends to roll out globally; piloted long-duration chemistries from Ambri and Primus Power.
• Meta: a 1 GW / 100 GWh ultra-long-duration supply agreement with Noon Energy; a 300 MW / 1,200 MWh Ørsted deployment supporting its Arizona site; a 200 MW / 1,600 MWh Tesla Megapack system paired with solar in Wyoming.

The common thread is vertical integration. Google, Microsoft, and Meta are not simply buying batteries. They are internalising storage procurement, which reshapes vendor relationships and financing models across the sector.

Vendor-side demand signal (unverified) Fluence has reported a global pipeline of 164 GWh, including approximately US$850 million in data centre business. As a vendor indicator of committed demand rather than pilot interest, treat the figure as directional and not independently confirmed.

Policy is moving in the same direction at the same time. FERC Orders 841 and 2222 opened wholesale capacity, energy, and ancillary services markets to behind-the-meter and aggregated storage, while Order 2023 restructured the interconnection queue. In June 2026, FERC issued Section 206 show-cause orders to six major grid operators, including PJM, MISO, and CAISO, directing them to clarify co-location rules and prevent cost-shifting (unverified).

Read the vertical integration moves as a structural shift in procurement power. The largest power buyers no longer trust the market to deliver storage at the scale and speed they need, so they are pulling that decision in-house, moving leverage away from utilities and toward the hyperscalers themselves. For positioning, the deployment map shows which vendors and configurations are earning real commitments, which is more useful than any forecast.

The barriers that will determine whether the next two years deliver

Now the friction. The barriers here are structural constraints with measurable dimensions, not vague risk factors, and the most underappreciated one reframes the entire thesis.

Early-life failure rate (unverified) Electric Power Research Institute (EPRI) data indicates roughly 72% of BESS failures occur during construction, commissioning, or the first two years of operation, with integration and assembly issues accounting for 36% of all failures.

That statistic shifts the question. The risk is not whether demand exists. It is whether deployments perform as modelled, and the failures cluster overwhelmingly in integration design, control systems, and dispatch governance rather than in the battery cells themselves. Which means the choice of integrator matters as much as the choice of cell chemistry.

The standardisation gap compounds the problem. The absence of standardised load profiles limits grid operator visibility into what data centres actually draw, which restricts power suppliers’ ability to meet demand and depresses the value BESS can deliver back to the network.

Physical footprint is a genuine competitive disadvantage. Large installations require up to 1,000 square feet per MWh (unverified), forcing storage to compete directly with revenue-generating IT space in land-constrained markets. That footprint connects to fire safety and permitting: even the safer lithium iron phosphate (LFP) chemistry, with a thermal-runaway threshold around 270°C against roughly 210°C for NMC (unverified), still demands rigorous suppression systems, and some local governments have imposed chemistry-specific restrictions or moratoria.

LFP chemistry trade-offs shape the permitting and fire-safety calculations the article references: higher thermal-runaway thresholds reduce suppression costs and ease local government approval, but LFP systems carry lower energy density than NMC, which directly worsens the footprint problem in land-constrained data centre markets.

What needs to change for the window to close in batteries’ favour

The Volta Foundation sets out four conditions, each aimed at a specific barrier:

1. Data centre developers and hyperscalers should specify hybrid architectures from the initial design phase, so batteries earn daily revenue and backup becomes a byproduct rather than the sole function. This directly counters the idle-cost-centre problem.
2. Grid operators and policymakers should standardise curtailment-based interconnection processes and clarify dispatch rates and large-load tariffs, which addresses the queue delays and investment confusion holding deployments back.
3. Battery manufacturers and integrators should differentiate products by load type, with high-power cells for volatile intra-day swings and dense energy cells for structural backup, resolving the one-size-fits-all mismatch.
4. The broader industry should collaborate to build a standardised data centre load profile the wider grid can design around, closing the visibility gap that depresses network value.

For investors, this is where value can be lost even inside a high-growth market. Integration quality, standardisation, and permitting friction are the variables that separate deployments that perform against modelled returns from those that quietly underdeliver.

Reading the competitive map as an investor

The synthesis is not a single verdict. It is a three-tier framework, and each tier calls for a different posture.

The winning segments warrant confidence: curtailment flexibility, UPS ride-through, sub-second power quality, and bridge power for queue-stalled assets are generating committed procurement now. The contested tier, sub-eight-hour backup economics and grid services, requires monitoring specific variables, chiefly regional demand-charge structures. The losing tier, multi-day and continuous prime power, still belongs to gas and diesel.

The global BESS demand trajectory matters for supply chain investors because hyperscaler procurement is only one source of pressure on cell and integrator capacity; utility-scale renewable firming and grid balancing commitments in Europe and Asia are absorbing similar lead times, which means the data centre window competes with broader market demand rather than sitting isolated from it.

The Volta Foundation frames the next two to three years as the decisive window in which the competitive balance is set. The gas turbine delivery constraint, with orders pushed into the next decade, is a near-term tailwind that is structural rather than cyclical.

Watch these variables as the window unfolds:

• Whether curtailment-based interconnection becomes standardised practice across grid operators.
• Whether long-duration storage cost economics keep improving toward the eight-hour threshold.
• How quickly interconnection queues actually clear.
• Whether gas turbine availability recovers ahead of schedule.

The thesis is conditional, and that is the point. If curtailment interconnection standardises and long-duration costs keep falling, batteries earn a permanent operational role in load shaping and grid services. If queues resolve quickly and gas availability increases, BESS reverts to idle backup, a materially weaker investment profile. The precise call depends on how a small number of identifiable policy and market variables resolve.

This article is for informational purposes only and should not be considered financial advice. Investors should conduct their own research and consult with financial professionals before making investment decisions. Past performance does not guarantee future results. Financial projections are subject to market conditions and various risk factors, and forward-looking statements are speculative and subject to change based on market developments.