Solid-state cooling without the gases
Magnotherm describes Polaris as the world’s first commercially available magnetic cooler (image: Magnotherm)
The heat generated inside an AI data centre is normally treated as waste twice over. Electricity powers the processors, then additional energy is used to move the resulting heat outside.
Liquid cooling can carry heat away more effectively than air, but the heat must still be transported and ultimately rejected. As AI racks become hotter and denser, thermal management is increasingly determining how much computing equipment operators can install and how much electricity can be allocated to revenue-generating processors.
Solid-state cooling proposes something more radical. Instead of compressing and expanding a refrigerant gas, it exploits materials that heat up and cool down when subjected to pressure, mechanical stress, magnetic fields or electric fields. Researchers are even investigating whether waste heat could provide some of the energy needed to drive the process.
If the technology can be scaled, cooling could become less of a one-way operational energy drain and more closely integrated with heat recovery and reuse. A data centre would still consume energy and reject heat, but some systems might eventually put part of an otherwise wasted thermal resource to work.
Why conventional cooling needs an alternative
When Jacob Perkins patented an apparatus for producing ice in 1834, he established the basic architecture behind most modern refrigeration.
A fluid circulates through a closed system. A compressor raises its pressure and temperature. The resulting heat is released elsewhere before the refrigerant expands, cools and absorbs more heat. Domestic refrigerators, supermarket freezers, air conditioners, heat pumps and industrial chillers still operate on variations of this vapour-compression cycle.
Nearly two centuries of engineering have made these systems remarkably effective. Compressors have become quieter, heat exchangers more efficient and control systems more precise. But the underlying machine remains dependent on moving parts, pressurised circuits and a refrigerant that repeatedly changes phase.
Compressing refrigerant consumes electricity, while gases escaping during manufacturing, maintenance or disposal can contribute directly to global warming. Some widely used hydrofluorocarbon refrigerants have a much greater warming effect than carbon dioxide. Lower-impact alternatives are increasingly available, but choices involving hydrocarbons, ammonia, carbon dioxide and newer fluorinated gases bring different trade-offs in flammability, toxicity, operating pressure, cost and performance.
Meanwhile, the world requires more cooling. The United Nations Environment Programme expects the installed capacity of cooling equipment to triple by 2050. Under current policies, electricity consumption for cooling would more than double and related emissions could reach 6.1 billion tonnes of carbon dioxide equivalent annually.
Several groups now think part of the answer could be to remove the gaseous refrigerant from the central cooling cycle.
How does solid-state cooling work?
Solid-state cooling exploits materials that change temperature when subjected to an external field or force. Apply that stimulus and the internal order of the material changes, releasing heat. Reverse it and the material can absorb heat from its surroundings.
The solid effectively performs the thermal role normally assigned to a refrigerant gas.
Different stimuli produce different branches of the technology:
- Magnetocaloric cooling uses magnetic fields.
- Electrocaloric cooling uses electric fields.
- Elastocaloric cooling places shape-memory alloys under mechanical stress.
- Barocaloric cooling exposes specialised materials to hydrostatic pressure.
The thermodynamics resemble conventional refrigeration: heat is collected in one place and discharged in another. The difference lies in how the temperature change is produced. Instead of compressing and expanding a gas, the system drives a reversible transition inside a solid.
This could eliminate fugitive emissions from high-global-warming-potential refrigerant gases. It does not necessarily eliminate all fluids, pumps or moving parts: several solid-state designs still circulate water or another heat-transfer medium between the caloric material and the places where heat is collected and released.
The science is not entirely new. Thermoelectric coolers, which create a temperature difference when current passes through semiconductor materials, are already used in electronics, medical devices and portable coolers. Their compactness and lack of moving parts are valuable, but their efficiency and heat-removal capacity have generally prevented them from displacing vapour compression in large refrigeration and air-conditioning systems.
Newer caloric systems promise higher performance, but have encountered other obstacles: expensive magnets, material fatigue, high forces or pressures and difficulty transferring heat rapidly enough to turn a momentary temperature change into continuous cooling.
Recent prototypes and early commercial products are beginning to reveal which of those obstacles can be overcome—and which may prevent particular approaches from scaling.
Companies developing solid-state cooling
German company Magnotherm has taken one of the first solid-state approaches into a product that customers can buy or rent. Its Polaris drinks cooler uses a lanthanum-iron-silicon magnetocaloric alloy and is marketed for supermarkets, events and shared spaces.
Inside Magnotherm’s system, the material heats when magnetised and cools when demagnetised. A mixture of water and alcohol flows through its porous structure, transferring heat to a radiator during one part of the cycle and carrying cooling to the drinks compartment during another.
Magnotherm describes Polaris as the world’s first commercially available magnetic cooler. That remains a company claim, but the product can be hired or purchased. Its modest application is instructive: beverage cooling provides a controlled environment in which Magnotherm can collect operating data before attempting much larger installations.
Cambridge University spinout Barocal is pursuing a different route. Its organic barocaloric materials change temperature when pressure alters their molecular structure. The company says its materials can deliver large thermal changes while remaining inexpensive and suitable for repeated operation, although independent system-level performance data will be needed to establish how they compare with commercial compressors.
Barocal raised $10 million in May 2026 to expand its engineering team and move towards commercial deployment. Its initial target markets include commercial refrigeration and data-centre cooling. No commercial data-centre installation has yet been disclosed.
US company Pascal is also developing barocaloric refrigeration. It is working to translate solid refrigerants from small proof-of-concept systems into a commercial-scale demonstrator. The company says its materials operate within temperature and pressure ranges comparable to those found in gaseous-refrigerant systems, while its recent industry presentations have emphasised the potential for low-pressure barocaloric refrigeration. These performance claims have not yet been validated in a commercial product.
Pascal raised $8 million in 2024 to support that scale-up. As of 2026, it remained in the development and demonstration stage rather than commercial production.
Qurie, founded in Germany in early 2026, represents the electrocaloric route. The Fraunhofer Institute for Physical Measurement Techniques spinout applies electric fields to ceramic components, causing their temperature to change.
Its patented design uses evaporation and condensation inside an active heat-pipe system to move heat rapidly away from the electrocaloric material. Fraunhofer reports that a research setup achieved a specific cooling or heating power of approximately 1,500 watts per kilogram of electrocaloric material. Qurie’s initial target markets are control-cabinet and laser cooling, with commercial refrigeration and consumer systems planned for later stages.
Together, these companies illustrate why solid-state cooling is not one technology. Each approach must balance material cost, efficiency, temperature range, heat transfer, mechanical complexity and durability. The eventual market may contain several caloric systems suited to different tasks rather than one universal replacement for the compressor.
Can elastocaloric cooling overcome material fatigue?
Elastocaloric cooling provides a useful example of the distance between an impressive material and a viable machine.
Shape-memory alloys release heat when mechanically stressed and cool when that stress is removed. They can produce substantial temperature changes, but repeatedly loading and unloading a material creates fatigue. A refrigerator expected to operate for years cannot depend on an alloy that gradually cracks or loses performance.
Researchers have now reported an elastocaloric device using a fatigue-resistant titanium-nickel-copper-cobalt alloy that maintained a cooling output of 400 watts and a temperature span of 41 kelvin over one million operating cycles. Its refrigerant structure survived more than ten million mechanical cycles in separate testing.
That is an important durability result, but one million cycles can still represent only days or weeks of operation in a device cycling several times per second. Commercial systems will require longer tests under realistic loads, alongside evidence on efficiency, manufacturing cost and maintenance.
A separate team at the Karlsruhe Institute of Technology and Japan’s University of Tsukuba has demonstrated a more experimental variation: an elastocaloric cooler powered by heat rather than an electric motor. One shape-memory film converts heat into movement; that movement loads and unloads another film to produce cooling.
With its actuator heated to 86°C, the prototype produced a four-degree temperature difference at component level. It also operated from an external heat source at 130°C. The published study reports milliwatt-scale cooling power, placing it far from commercial refrigeration or data-centre deployment.
Its significance lies in the mechanism. Waste heat or solar thermal energy could, under suitable conditions, replace some of the electrical actuation normally required by an elastocaloric system.
Could solid-state cooling reduce data-centre energy use?
AI is making thermal management harder because more computing power is being concentrated inside individual racks. AI installations commonly reach tens of kilowatts per rack, while newer designs are moving into hundreds of kilowatts. Air cooling is becoming impractical at the highest densities, accelerating the adoption of direct-to-chip liquid cooling.
Solid-state cooling addresses a different part of the system. Liquid carries heat away from processors; caloric technology could potentially change how the required temperature difference is generated.
For a data-centre operator, the possible gains extend beyond compressor electricity. Solid-state systems would avoid gaseous-refrigerant emissions and could potentially target processors, optical components or other hotspots rather than cooling an entire room around its hottest equipment.
The most ambitious possibility is a partially regenerative thermal system. Heat collected from processors could be used for district heating, buildings or industrial processes and, if available at a sufficiently high temperature, might also provide some of the driving energy for caloric cooling.
That final condition is substantial. The Karlsruhe–Tsukuba experiment used an actuator temperature of 86°C and demonstrated operation with a 130°C source. Data-centre cooling loops often deliver lower-grade heat. Converting the laboratory principle into a self-assisted data-centre cooling system would require different materials, higher waste-heat temperatures or an additional heat-upgrading stage.
Cooling would therefore remain an operational expense. Thermodynamics does not permit a perfectly circular refrigeration system, and heat still has to leave the facility. The realistic opportunity is to extract more useful work from the energy entering the data centre before it is discharged.
Barocal explicitly identifies data centres as an initial target. The attraction is understandable: operators evaluate cooling according to lifetime cost, reliability and the amount of site power left for computing. Even a relatively expensive new system could find a market if it lowered energy use or enabled more processors to operate within an existing power envelope.
The hurdle is scale. A drinks cooler and an AI rack occupy different engineering worlds. Data-centre equipment must remove large and rapidly changing heat loads continuously, tolerate component failures and meet exacting uptime requirements. Solid-state systems will also compete with direct-to-chip liquid cooling and warm-water, chiller-free designs that are already moving into deployment.
What is preventing solid-state cooling from scaling?
Vapour-compression equipment benefits from enormous manufacturing scale, established supply chains and generations of practical experience. Engineers understand how it behaves across tropical heat, domestic neglect, industrial vibration and years of continuous operation.
Natural refrigerants such as carbon dioxide, ammonia, propane and isobutane can reduce climate impact without replacing the entire architecture. Solid-state systems must therefore demonstrate not only attractive material-level results, but competitive whole-system efficiency, purchase price, temperature range and reliability.
They still require heat exchangers, controls and, depending on the approach, pumps, magnets, high-voltage electronics or pressure-generating equipment. Removing the compressor does not automatically remove mechanical or operational complexity.
The first successes are consequently likely to appear in tightly defined markets: drinks coolers, medical refrigeration, electronics, control cabinets, lasers or specialised industrial equipment. These applications may tolerate higher initial costs when compactness, low noise, precise temperature control or the absence of gaseous refrigerants provides a particular advantage.
From there, manufacturers can accumulate the operating evidence required for larger systems.
The compressor is no longer competing only with incremental improvements to the refrigerator. It is being tested against an emerging model of thermal infrastructure in which heat is captured, moved, reused and, perhaps eventually, recruited to help drive cooling.
Data centres offer the largest prize and the harshest test. If caloric systems can move heat at rack-scale power densities while demonstrating competitive efficiency and years of reliable cycling, operators could recover more value from every unit of electricity entering the facility.
Cooling would not become free or thermodynamically circular. But for an industry spending heavily to generate heat and then spending again to remove it, even a tighter energy loop could rewrite a significant part of the operating equation.
Liked this article? You can support our independent journalism via our page on Buy Me a Coffee. It helps keep MoveTheNeedle.news focused on depth, not clicks.