Thermal Storage:
The Overlooked Half

Ask an owner how they plan to shift load off the afternoon peak and the answer is almost always a battery. It is a reasonable answer. It is also, for many industrial and commercial sites, an answer to a narrower question than the one that matters, because a large part of the demand being shifted is not electricity in any useful sense. It is cooling, refrigeration, hot water, steam, or process heat, and it only looks like electricity because that is where the meter sits.

Storage that holds the thermal service directly, whether a tank of chilled water, a bank of ice, a hot-water or steam accumulator, or a mass of heated solid media, is a different asset with a different cost structure, a different failure mode, and a different set of things it cannot do. It is not better than a battery. It is not worse. It is a separate candidate that many site-level energy conversations never formally screen, and a decision that skips it is a decision made on an incomplete list.

This paper sets out where the thermal share of electric demand actually sits, the mechanism by which thermal storage saves money, the honest case against it, how it compares with electrochemical storage and with generation, and one detail in the federal tax definition that surprises owners late in a project.

Section 01The load behind the load

The starting fact is not about storage at all. It is about what the electricity is being used for.

The U.S. Department of Energy, in its 2023 technology strategy assessment on thermal energy storage, states that about 74 percent of total U.S. electricity is consumed in buildings, and that more than 50 percent of that consumed electricity goes to meeting thermal demands: space heating, space cooling, water heating, and refrigeration.¹ The Energy Information Administration's most recent commercial buildings survey puts cooling alone at about 14 percent of commercial building electricity consumption, roughly 170 billion kilowatt-hours.²

Those are national aggregates, and no single site is the aggregate. A machine shop is mostly motors. A data hall is mostly servers, though its cooling share is larger than most owners assume. A refrigerated warehouse is overwhelmingly a refrigeration plant with a building wrapped around it. A food processor, a brewery, a commercial laundry, a hospital, a distribution center with cold rooms. Each carries a thermal fraction that can be measured rather than guessed, and the measurement changes the shape of the storage question.

The reason this matters for load shifting is mechanical. A battery shifts electricity regardless of what the electricity is for; it is service-agnostic, which is its great strength. Thermal storage shifts exactly one service and nothing else. If the thermal fraction of a site's coincident peak is small, thermal storage is close to irrelevant no matter how attractive its economics look in isolation. If that fraction is large, the site has a second, quieter option that will not appear on any storage quotation, because nobody sells it as storage.

Section 02What thermal storage actually is

DOE groups thermal storage into three formats.¹ Sensible storage changes the temperature of a medium, whether water, oil, rock, sand, concrete, or molten salt, and is the format behind almost everything commercially deployed today. Latent storage uses a phase change, most familiarly water to ice, which packs more energy into less volume because melting absorbs considerably more energy than warming. Thermochemical storage holds energy in a reversible chemical reaction and remains largely a research category.

In practice, the commercial forms an owner will encounter are narrow and old. Chilled-water tanks, stratified so cold water is drawn from the bottom and returned warm to the top. Ice storage, built at night and melted through the afternoon. Hot-water tanks and steam accumulators, which have been standard equipment in industrial plants for a century. Heated solid media for high-temperature process heat, which is the newest and least mature of the group. And the building or the stored product itself, pre-cooled or pre-heated so that its own mass carries the load through the expensive hours.

Maturity deserves a clear statement, because the marketing around energy storage tends to blur it. DOE notes that nearly 85 percent of installed thermal storage capacity serves building, district, and industrial process heating.¹ Storage that takes in heat or cold and gives back heat or cold is not novel; it is unglamorous infrastructure with a long service record. The genuinely emerging category is the bidirectional kind: electricity in, heat stored, electricity out. That version requires a power cycle on the discharge end and carries the conversion losses that come with one. DOE's long-duration storage target frames that challenge plainly: a 90 percent cost reduction for technologies delivering ten hours or more of storage within the decade.¹ That is a research goal, not a market price, and it should not be read as a forecast of what an owner can buy this year.

Section 03Why the arithmetic can favor thermal

The core advantage is that thermal storage never converts back.

An electrochemical system takes electricity in and gives electricity out, and the round trip has a cost. PNNL's grid analysis documentation, which defines round-trip efficiency as the ratio of total energy out to total energy in measured at the point of connection, carries a median lithium-ion value of 83 percent drawn from the storage literature it cites.³ Installed performance varies with temperature, duty cycle, and how the balance of plant is counted, and any specific system should be evaluated on its own measured data rather than on a literature median. The direction, though, is not in dispute: some fraction of what goes in does not come out.

Thermal storage stores the service instead of the electricity. Ice made at two in the morning is cooling, and it is still cooling at four in the afternoon. There is no second conversion, only standby loss through the tank wall. DOE makes the same point in the industrial context, observing that avoiding the efficiency loss of electricity generation is one reason process-heat applications have attracted the thermal storage industry's attention.¹

A second, smaller effect works in the same direction for cooling systems. Chillers reject heat to ambient air, and ambient air is cooler at night. A machine making chilled water at two in the morning is generally rejecting heat against a lower condensing temperature than the same machine at four in the afternoon, so it consumes less electricity per unit of cooling delivered. The effect is real, it is site-specific, and it is regularly overstated. It also runs against a countervailing penalty when the storage medium is ice rather than chilled water: ice requires a lower evaporator temperature than chilled water, which cuts the other way and can raise electricity consumed per unit of cooling. Whether the night-time gain or the ice-making penalty dominates is a question of equipment, climate, and control strategy, to be settled with measured data on the specific plant rather than assumed in either direction.

The third effect usually decides the economics, and it has nothing to do with efficiency. It is the tariff. PG&E's commercial time-of-use plans place the on-peak period at 4 to 9 p.m. every day, with the summer season running June through September and partial-peak periods bracketing the peak window.⁴ That is a narrow, predictable, entirely foreseeable five hours. An asset that has only to carry a thermal load across a known window, on known days, in a known season, is a much easier engineering problem than an asset that must be ready for anything. Storage sized to a published tariff window is storage sized to a fact rather than to a scenario.

Section 04The case against, stated plainly

Every advantage above has a matching cost, and an owner who hears only the first half is being sold to rather than advised.

Shifting load can raise total energy use

This is the finding owners least expect. Lawrence Berkeley National Laboratory's strategy guide for refrigerated warehouses, prepared for the Department of Energy, modeled pre-cooling as a load-shifting measure and found it effective at holding space temperature during an event when refrigeration is off. It also found that pre-cooling increased overall refrigeration system energy use by approximately 5 percent on a given event day, because evaporator coil fan speed rose during the pre-cooling period to accomplish the over-cooling, and the added fan energy outweighed the compressor savings from operating at cooler night-time conditions.⁵

The same report is careful about the boundary of that finding: on systems without evaporator fan speed control, the fan energy increase would not apply, and pre-cooling would likely produce a slight net energy saving.⁵ That is the whole lesson in two sentences. The penalty is not a property of thermal storage; it is a property of a specific plant configuration. It has to be modeled on the actual equipment, and a study that reports demand savings without reporting the energy consequence has told half the story.

It only moves one service

A chilled-water tank does not shift a compressor line, a kiln, a fleet of chargers, or a server hall's IT load. If the site's coincident peak is set by something other than the thermal service, thermal storage can run perfectly and change the bill very little. The screening question is not whether the site has a large cooling load. It is whether the cooling load is on when the peak is set.

The process may not tolerate the excursion

Pre-cooling assumes the temperature can move. Often it cannot. The LBNL guide notes that frozen product storage is generally a good candidate because product can be cooled somewhat without affecting quality, while spaces held around 30 to 35 degrees Fahrenheit present a real problem, since the goal there is to keep product as cold as possible without freezing it, so pre-cooling is limited or infeasible. It further lists produce vulnerable to chilling injury that should not be over-cooled, among them cucumbers, cranberries, eggplant, melons, okra, pumpkins, squash, white potatoes, sweet potatoes, and tomatoes.⁵ The equivalent constraint exists in process heat, where product quality and metallurgy set the temperature band, not the tariff.

It takes space, and space has a cost

Thermal storage is volumetric. A chilled-water tank is a large object that must go somewhere, on a foundation, within a fire and seismic design, in a yard that may already be committed to trucks. Ice reduces the volume relative to chilled water but does not eliminate the problem. On a constrained urban site, the space argument alone can end the conversation, and it should be tested before any modeling begins rather than after.

It cannot earn revenue the way an electrical asset can

A battery can be dispatched for more than one purpose. Thermal storage generally serves the host process and nothing else. It does not export, it does not respond to a wholesale price signal on its own, and it typically has fewer routes into grid programs. Where an owner is counting on stacked value streams to justify capital, that difference is decisive.

Somebody has to run it

The savings come from a control strategy executed every day: charge overnight, hold, discharge across the peak window, adjust for weather and for production schedules. That strategy decays quietly when the person who understood it leaves, and the asset keeps operating in a degraded mode that nobody notices because nothing breaks. Any storage business case that depends on daily operating discipline should say who owns that discipline and what happens when they move on.

Section 05Comparing the options without a favorite

The honest framing is not thermal versus electrochemical. It is a list of candidates, each of which shifts something specific, needs something specific, and fails to do certain things at all.

Two entries in that table are worth dwelling on because they are the ones owners skip. The last row, changing the tariff or the schedule, costs nothing and is the correct answer more often than the storage industry finds convenient. And the fourth row, the mass already sitting on site, is free thermal storage that most facilities have never characterized. A cold warehouse full of frozen product is a thermal battery whether or not anyone accounts for it. That same mass is why a refrigerated facility can ride through a short outage without spoilage, a resilience property owners often pay a second time to buy in another form.

Section 06Where it changes the time-to-power answer

For sites facing a multi-year wait for new or expanded grid capacity, the more interesting use of thermal storage is not the bill. It is the size of the request.

A utility service is sized to peak demand, not to annual energy. If a meaningful share of the peak is thermal, and if that thermal load can be built overnight and drawn down through the afternoon, then the peak the utility is asked to serve may be materially smaller than the peak the site would otherwise present. Whether that reduction is large enough to change the service required, keep a project within existing capacity, or alter the queue position is a site-specific engineering and tariff question with no general answer. It is, however, a question worth asking before an upgrade application is filed, because an application is a slow thing to revise.

The same logic applies to a project that intends to build generation. Thermal storage that flattens the thermal peak reduces the capacity that any other asset has to cover. It is not a substitute for generation and it does not deliver power during an outage. It changes the number that generation, or a battery, or a service upgrade is sized against, and everything downstream of that number gets cheaper.

Section 07Where the tax definition stops

This is the detail that surprises owners late, and it is worth reading the statute rather than a summary of it.

Federal law includes thermal energy storage property within the definition of energy storage technology for the investment credit. But the statutory definition is specific: thermal energy storage property means a system directly connected to a heating, ventilation, or air conditioning system, which removes heat from or adds heat to a storage medium for subsequent use, and which provides energy for the heating or cooling of the interior of a residential or commercial building. It expressly excludes a swimming pool, combined heat and power system property, and a building or its structural components. The separate electricity-storage branch of the same definition carries a minimum nameplate capacity of not less than 5 kilowatt-hours.⁶ The clean electricity investment credit adopts that same definition by cross-reference.⁷

Read carefully, the definition is tethered to conditioning the interior of a building. An ice tank serving a building's air conditioning sits comfortably inside it. A heated media store charging overnight to feed a process oven is doing something the plain words of that definition do not obviously describe, and the exclusion of a building and its structural components means a design that stores heat in the structure itself invites the same question. Whether a given configuration qualifies, and under which provision, is a determination for qualified tax counsel on the specific facts. It is not an assumption a feasibility model is entitled to make.

On the credit rate: under current law, qualifying energy property is credited at 30 percent, with the statute setting a lower base rate and the higher rate available where its stated conditions are met.^6,7 This paper assumes no adders. Statutory values are stated as of August 2026 and change through legislation, regulation, and guidance; confirm the current position with qualified tax counsel before any figure enters a capital plan.

Section 08Five questions before thermal storage enters a plan

1. What share of the coincident peak is thermal?Not annual thermal energy. The thermal load present in the hours that set the demand charge and the service size. This comes from interval data and submetering, and it is measured, not estimated.
2. How far can the temperature move, and for how long?Product limits, process limits, comfort limits, and code limits define the storage capacity that is actually usable. A tank sized past that boundary is a tank that never fully discharges.
3. What does the shift cost in energy?Model the total kilowatt-hours before and after, on the real equipment and its control capability. If total consumption rises, that cost belongs in the same table as the demand savings, not in a footnote.
4. Who operates the strategy, and what happens when they leave?Daily savings depend on daily discipline. Name the role, document the sequence, and assume turnover.
5. Does it change what you must ask the utility for?If flattening the thermal peak reduces the service capacity required, that may be worth more than the tariff savings. Test it before the application is filed.

Section 09What this means for a decision

Thermal storage is not an answer. It is a candidate that belongs on the list alongside batteries, generation, tariff changes, load flexibility, and doing nothing at all. It wins where the thermal share of the peak is large, the process tolerates a temperature excursion, space exists, and the site has the operating discipline to run it. It loses where the peak is set by non-thermal load, where the product cannot move a degree, where the yard is full, or where the case depends on revenue streams a thermal asset cannot reach.

What makes it worth a paper is that it is systematically under-screened. Nobody arrives at a site to sell a chilled-water tank the way they arrive to sell a battery or a generator, so the option is often absent from the comparison rather than rejected by it. An honest decision process fixes that by putting every path on the same page, with the same sourcing standard, and letting the site's own measured load decide which one survives.

That is the whole method: measure first, list every candidate, price the trade-offs each one actually carries, and be equally willing to recommend the tank, the battery, the engine, the tariff change, or nothing.