Summary
- KIT and the University of Tsukuba have built a prototype in which one shape-memory film drives another to produce cooling.
- The feasibility study is not yet designed for useful data centre-scale cooling capacity.
- Researchers are working on parallelising multiple films to increase cooling output and explore processor-level applications.
Karlsruhe Institute of Technology and the University of Tsukuba have demonstrated a heat-driven solid-state cooling prototype that replaces the electric motor or compressor used in conventional refrigeration with interacting shape-memory films.
The research uses two ultra-thin nickel-titanium foils. One responds to heat by generating mechanical movement; that movement acts on the second film, which produces an elastocaloric cooling effect as it is mechanically loaded and unloaded.
The result is a system in which heat provides the driving energy for the cooling cycle. KIT says the prototype is a feasibility study rather than an optimised cooling machine, and the researchers are now working on connecting multiple films in parallel to increase cooling capacity.
That distinction is essential for data centre applications. The system demonstrates a physical mechanism that could eventually be useful around electronics, but it does not yet displace chillers, cooling towers, coolant distribution units, or direct liquid-cooling loops at facility scale.
Cooling without a compressor
Most mechanical refrigeration systems move heat by compressing and expanding a refrigerant. The compressor consumes electricity and contains moving mechanical components, making it one of the major energy users in many cooling systems.
Elastocaloric materials offer another route. Shape-memory alloys change temperature when mechanical stress is applied and released. By controlling that cycle, heat can be moved without a conventional vapour-compression refrigerant circuit.
The KIT and Tsukuba researchers have taken that principle a step further by using heat itself to produce the mechanical work required to drive the cooling film. Potential heat sources include solar thermal energy or waste heat already produced by the equipment being cooled.
That creates an intriguing feedback possibility for computing: a processor produces heat, and part of that thermal energy could theoretically help drive a local cooling mechanism. KIT specifically identifies processor cooling as a potential future application.
The engineering gap between that concept and a data centre is substantial. Modern AI servers can dissipate kilowatts per chassis and tens or hundreds of kilowatts per rack. A practical cooling system has to move that heat continuously, with predictable performance across changing loads, while meeting stringent reliability and maintainability requirements.
The research team has not claimed that the current prototype can do that. Its immediate work is focused on scaling output by arranging multiple shape-memory films in parallel.
Why solid-state cooling attracts attention
Cooling power has become more important to data centre economics as rack densities rise. Direct-to-chip liquid cooling can remove heat efficiently from high-power processors, but the heat still has to travel through coolant loops, heat exchangers, pumps, and ultimately a facility heat-rejection system.
Solid-state technologies are attractive because they could reduce or remove some mechanical components and refrigerants. In principle, they can also be compact enough to place cooling closer to the heat source.
Reliability will be a major issue. Shape-memory materials undergo repeated mechanical phase changes, so a deployable system would need to demonstrate long cycle life, stable performance, manufacturability, and fault tolerance. It would also have to show that any efficiency advantage survives once pumps, controls, heat exchangers, and supporting equipment are included.
The work, published in Nature Energy, is therefore better viewed as a new thermal-management building block than a near-term replacement for established data centre cooling infrastructure.
Its relevance comes from the direction of travel. Higher-density computing is forcing engineers to reconsider where cooling happens and which energy sources drive it. A technology capable of turning otherwise unwanted heat into useful cooling would alter that balance, particularly at component or server scale.
The next evidence will come from capacity, cycle-life, and integration testing. Until then, the prototype shows that the thermodynamic mechanism can work; it does not yet show that it can carry the thermal load or uptime burden of a production data centre.
