Episode 58: Can Batteries Really Power the Data Center Boom?

The following is the transcript of Episode 58 of The Solar Coaster Podcast. This episode features AI-generated voice narration of Anna Covert and Alex Herrera.

Full Episode Transcript

Anna Covert: Welcome to the debate. Data centers are expanding rapidly, driven in large part by artificial intelligence, and that expansion is creating a difficult power question: where can battery energy storage systems actually compete? The answer is more limited, and more interesting, than the usual claim that batteries can solve everything. A recent analysis from the Volta Foundation divides data center applications into three groups. Batteries are winning in two areas, competing in four, and losing to other technologies in two more. So today, the question is not whether batteries belong in the data center. They already do. The question is where they create real value, where their economics remain conditional, and where other technologies are still better suited.

Alex Herrera: And I would start by challenging the headline that batteries are simply the future of data center power. That is too broad. Batteries are very good at responding quickly, shifting power over short periods, and managing peaks. But data centers also need sustained energy, long-duration backup, and continuous prime power. Those are very different requirements. A battery that can respond in milliseconds is not automatically a battery that can keep a facility running for many hours, or supply continuous power day after day. The technology has to be matched to the job. The Volta Foundation's central argument is essentially that flexibility is the battery's main advantage, not unlimited endurance.

Anna Covert: The clearest wins are peak shaving and bridge power. Peak shaving means discharging the battery during costly periods in order to reduce demand charges. For a data center, that can be a meaningful economic benefit because the battery is helping control the facility's highest power draw. But even this advantage depends on the local electricity tariff. If a region does not impose demand charges, the strongest economic case for peak shaving can disappear. That is an important qualification. Batteries do not create the same value in every market simply because the hardware is the same.

Alex Herrera: Exactly. The battery is not valuable in isolation; it is valuable within a particular rate structure and operating profile. Developers need to ask when the facility peaks, how long those peaks last, what the tariff charges for them, and whether the battery can cycle often enough to justify its cost. A battery may look attractive in a market with substantial demand charges and much less attractive in a market without them. So when people quote a general case for battery storage at data centers, I would want to see the local tariff before accepting the conclusion. The economics are not universal.

Anna Covert: The second clear win is bridge power. Data center projects can spend a long time waiting in an interconnection queue. A battery can provide temporary power while the facility waits for a more permanent connection or generation arrangement. That is the speed-to-power advantage. In a market where time determines whether a project can proceed, getting some power sooner may be worth more than choosing the option with the lowest long-term cost. This is why battery suppliers and data center developers are treating storage as a practical part of the power strategy, rather than only as a backup device.

Alex Herrera: But even bridge power needs a correction. Standalone batteries rarely carry that role alone for a long period. The Volta analysis identifies fuel cells as a competitor for bridge power, and that comparison matters because bridge power may last longer than a short battery discharge. A battery can respond immediately, but it may need another source behind it. So the likely architecture is often hybrid: the battery handles rapid response and short-term transitions, while gas engines, turbines, fuel cells, or another source supplies longer-duration energy. Calling the battery the bridge does not mean the battery is the entire bridge.

Anna Covert: That hybrid idea is central. The report says batteries earn a place alongside firm generation. In other words, storage does not necessarily replace generation; it adds a flexible power-management layer. A battery can absorb or deliver power quickly, reduce peaks, support transitions, and provide operational flexibility while a firm source handles sustained output. This can be especially relevant as gas turbine supply constraints and changing power architectures affect project decisions. If a conventional generation option cannot arrive quickly enough, storage becomes more attractive even when it is not the cheapest source for every hour.

Alex Herrera: Still, we should not let urgency erase engineering limits. Batteries are competing, rather than winning outright, in four additional areas: curtailment flexibility, uninterruptible power supply ride-through, grid services, and sub-second power quality. These are meaningful applications. Software and power electronics have improved enough for batteries to participate in them. But competition means the result depends on system design, controls, market rules, and the value assigned to the service. A technically capable battery does not automatically receive payment or operational priority for every service it can provide.

Anna Covert: The areas where batteries lose are even clearer. Long-duration backup beyond four hours is still generally better served by diesel, according to the analysis. Continuous prime power is primarily served by gas turbines and fuel cells. This does not mean battery technology cannot improve, or that every site will use exactly those technologies. It means that current battery systems are not automatically competitive when the requirement is sustained energy over a long period or continuous operation as the primary power source. The duration requirement changes the economics and the physical design.

Alex Herrera: That distinction should be kept in front of every procurement discussion. High power and high energy are not the same thing. A data center may need a system that can deliver a large amount of power instantly, or it may need a large amount of stored energy over many hours. Those requirements affect the cells, the system size, the controls, the cost, and the operating strategy. The report recommends that battery and integration companies develop products around load archetypes. Some loads call for high-power cells; others call for energy-dense cells. Treating all data centers as one identical customer is a design mistake.

Anna Covert: There is also a deployment problem that has received less attention than the cell itself. The report says issues often emerge during commissioning and the first two years of operation, with integration, controls, and dispatch governance identified as the biggest culprits rather than cell chemistry. That is a strong warning for buyers. Choosing a battery with impressive specifications is not enough. The system has to work with the facility's electrical architecture, control software, operating procedures, and market obligations. The handoff between the battery, the generation assets, the grid connection, and the data center load has to be designed from the beginning.

Alex Herrera: And the industry still lacks a clearly defined method for integrating batteries into data centers. That uncertainty creates risk for investors and operators. There is also a lack of visibility into actual data center load profiles, which makes it harder for power suppliers to design the right solution. An artificial intelligence facility may have a different demand pattern from another data center, and the same facility may change as its computing workload changes. Without a detailed understanding of when power is needed and how quickly it changes, developers may oversize the battery, undersize it, or assign it a role it cannot reliably perform.

Anna Covert: That is why the recommendation to design hybrid architectures from day one is so practical. Developers and hyperscalers should specify the battery for daily duty, so that backup capability arrives as a byproduct of a system already being used. This approach can give the battery a regular operational role in load shaping, peak reduction, and possibly grid services, instead of leaving an expensive asset idle while waiting for a rare outage. It also connects the power decision to time-to-power. The report recommends ranking power options on how quickly they can be deployed alongside their levelized cost.

Alex Herrera: I agree with the logic, but daily cycling also introduces discipline. If the battery is expected to shave peaks, provide grid services, support ride-through, and remain ready for backup, those duties can conflict. The operator needs dispatch governance: which service has priority, what state of charge must be maintained, and who is authorized to change the operating strategy? These are not secondary software details. They determine whether the battery is dependable when the facility needs it. A system that earns revenue by cycling aggressively may not be prepared for an unexpected outage unless the rules are explicit.

Anna Covert: The next two or three years may determine whether batteries become a permanent part of data center operations or are remembered mainly as a temporary bridge while interconnection queues remain slow. If curtailment-based interconnection becomes common and duration economics continue improving, installed batteries could take a standing role for the life of a facility. They could shape load, shave peaks, and sell grid services. But if interconnection queues clear faster than expected and gas supply becomes more available, some batteries may revert to a more limited backup role. The outcome depends on both technology and market conditions.

Alex Herrera: That uncertainty is why the strongest argument for batteries may not depend on arbitrage economics at all. Even when buying electricity at one time and using it later is not especially profitable, a battery can still provide value as the facility's power-management layer. It can respond quickly, manage transitions, and coordinate different sources. But that value is only realized if the industry understands the load and builds the controls correctly. The phrase power-management layer sounds attractive, yet it must translate into a defined operating function, measurable performance, and a clear responsibility for dispatch.

Anna Covert: The buyer landscape could also shape the technology. Large hyperscalers, including companies such as Google and xAI, appear to prefer vertically integrating their storage. If that pattern continues and data center operators become the largest buyers of grid-scale batteries, system specifications and financing models could change. The United States is expected to be the largest data center market for battery energy storage, and the way those loads connect to the grid may influence practices elsewhere. A major buyer with a strong preference for integration can push suppliers toward more customized systems.

Alex Herrera: But vertical integration has its own question: does it improve accountability, or does it concentrate complexity inside the operator? A data center company may understand its load better than an outside developer, but it still needs expertise in batteries, controls, grid participation, and safety. The more customized the system becomes, the more important standardized interfaces and operating rules may be. Integration can produce a better fit, but it does not remove the need for coordination with grid operators, generation suppliers, and equipment manufacturers.

Anna Covert: So where does that leave the debate? Batteries clearly win when speed, flexibility, and daily cycling matter. They are strong for peak shaving where demand charges support the economics, and they can help provide bridge power while projects wait for interconnection. They are credible competitors for curtailment management, UPS ride-through, grid services, and fast power-quality support. Those are substantial roles, especially for facilities whose economic value depends on uptime and rapid deployment.

Alex Herrera: And they lose, at least for now, when the central requirement is sustained energy, long-duration backup, or continuous prime power. Diesel, gas turbines, and fuel cells remain important competitors in those applications. The correct conclusion is not that batteries are inadequate, nor that they replace every conventional source. The conclusion is that batteries should be specified according to the load, the tariff, the interconnection rules, and the operating priorities. A battery can be the right answer for one part of a data center's power system and the wrong answer for another.

Anna Covert: The practical message to developers is to make those distinctions early. Define the load profile, design the hybrid architecture from the start, specify whether the battery is optimized for power or energy, and decide how backup, daily operation, and grid services will interact. Also, value time-to-power explicitly. A solution with a lower levelized cost may not be useful if it cannot arrive when the project needs it. For policymakers and grid operators, the message is to standardize curtailment-based interconnection, clarify dispatch and market participation, and create large-load tariffs that reduce confusion and delay.

Alex Herrera: For the battery and integration industry, the message is equally direct: move higher up the stack. Do not sell only cells or containers. Build systems around recognizable data center load patterns, with controls, integration, and dispatch governance treated as core products. The future of batteries in data centers will not be decided by capacity alone. It will be decided by whether the system can deliver the exact service promised, within the exact power architecture, under the exact market rules that apply to the facility.

Anna Covert: Then the balanced verdict is this: batteries are neither a universal replacement for firm generation nor merely emergency equipment. They are flexible infrastructure. Their best opportunities are the jobs that reward fast response and repeated daily cycling. Their weaker opportunities are the jobs that demand long endurance and continuous output. The market is still being defined by regulators, developers, hyperscalers, and suppliers, and the choices made now will determine whether storage becomes a permanent operating layer or a temporary solution to a temporary bottleneck.

Alex Herrera: And that is the question worth carrying forward. When a new data center asks for power, the first question should not be, “Can batteries do it?” It should be, “Which part of the power problem rewards what batteries do best?” If the answer is rapid response, peak management, bridge capacity, or flexible coordination, batteries may be highly competitive. If the answer is many hours of sustained backup or continuous prime power, another technology may be better. The future will probably belong to combinations, with each resource assigned the job it can perform most reliably. That is a more cautious conclusion than a battery-only vision, but it is also a more defensible one.