THE data centers they are the invisible heart of our digital life, but also enormous energy guzzlers. Every online search, every streaming video, every artificial intelligence model needs servers working non-stop and, above all, increasingly powerful cooling systems. Now, however, scientific research opens up an unprecedented scenario: bring a fluid below zero in less than 30 secondswithout using classic compressors. A perspective that could reduce consumption, costs and emissions.
The study, published in the prestigious journal Natureis already at the center of the debate between engineers, environmentalists and big tech. And while someone – how Elon Musk – imagine data centers in orbit, the real revolution could start much closer to us, from cooling systems.
When we think of cold, we immediately imagine a refrigerator or an air conditioner. In reality these systems all work the same way: a compressor circulates a refrigerant gas which absorbs heat slowly and releases it outside. It is an effective process, but energy-intensive and inflexible.
The technique described above Nature instead it follows another path. Here there is no mechanical compression: cooling occurs thanks to ultra-fast phase transitions or to details induced caloric effects. In simple words, by applying an external stimulus – such as an electric field or a controlled variation in pressure – the molecules of the fluid rearrange themselves instantlycausing the temperature to drop in a few seconds.
It is therefore not a question of “cooling more”, but of cool betteracting directly on the structure of the fluid. It’s this difference that makes the technology so attractive for data centers.
From laboratories to data centers
There is only one key question: Can it really be used in large systems? In laboratories the system works very well, but a data center is not a controlled experiment. We are talking about environments that work 24 hours a day, with constantly changing heat loads.
According to the researchers, scalability is possible, but requires custom-designed systems. Ultra-rapid cooling only works if the physical stimulus is distributed evenly. This is why they imagine themselves modular heat exchangerscapable of replicating the same effect on large volumes of fluid.
The advantage, however, is evident: compared to traditional systems, this technology could react almost in real time to heat peaks generated by servers and artificial intelligence chips, avoiding wasted energy.
Less energy, less emissions
Today, a huge portion of the energy consumed by data centers serves only to keep them cool. Reducing this requirement would mean lower the costs of digital servicesbut also reduce environmental impact of the internet and AI.
The benefit would be double. On the one hand less electricity consumedon the other the possibility of reduce or eliminate traditional refrigerant gasesmany of which have a very powerful climate-altering effect, much greater than CO₂.
The most realistic scenarios talk about one reduction in consumption for cooling between 20 and 30% in large data centers. This is no small thing, if we think about the continuous growth of data traffic and artificial intelligence.
Risks not to be underestimated
It’s not all rosy. The study also highlights very concrete problems. The use of fluids with salts or additives can cause corrosion and scalingwhile the very rapid cooling cycles subject pipes and valves to strong mechanical stress.
This is why they study stronger metal alloysadvanced protective coatings and systems intelligent monitoring. Digital sensors could warn of material degradation in advance, avoiding sudden and costly failures.
The idea of data centers in orbitevaluated by reality as SpaceX And xAI. The idea is fascinating, but hides a paradox: It is more difficult to cool in space than on Earth. In a vacuum there is no air and therefore no convection. Heat can only be dissipated by radiation, through huge radiators. In short, the “cold of space” does not automatically solve the problem.
In the global race for AI, among models such as ChatGPT and giants like OpenAI, Google And Halfit’s not just those with the most powerful chips who win. Without efficient cooling, even the most advanced processor must slow down. That’s why this technology could become a key piece to making AI more sustainablereducing the energy burden of a continuously expanding sector.
