Researchers have discovered a cooling method that can significantly reduce temperatures within seconds. This cooling could be the solution to many problems associated with the AI hype: the high energy consumption and the large waste heat of data centers.
What have the experts found? Researchers from the Chinese Academy of Sciences have found that water saturated with ammonium thiocyanate salt, when pressurized and then suddenly depressurized, cools down by 30 °C at room temperature and can even cool by up to 50 °C in warmer environments. You can find the entire research paper directly at Nature.com.
What advantages does the new cooling method have? In addition to its very rapid cooling performance, there are mainly two points that are advantageous:
- On the one hand, the theoretical efficiency is supposed to be much higher than that of conventional cooling methods currently in use.
- On the other hand, no special construction measures are required: Even though ammonium thiocyanate is a salt, it does not attack many common metals, so no special piping is necessary for the cooling.
What other problems still exist? The cooling still requires electricity to compress the refrigerant. The salt can also cause burns and irritations, making it harder to handle than some other compounds.
This novel cooling method is especially interesting for data centers
Who is the new cooling method interesting for? This novel cooling method is particularly exciting for modern data centers, of which there are increasingly more due to the AI hype. However, all of them share a common problem: The energy consumption of data centers is enormous and the waste heat from servers must also go somewhere. Without cooling, servers cannot operate and shut down.
Here, the novel cooling method could solve several problems at once: on the one hand, significantly reducing energy consumption and on the other hand, lowering the temperatures.
Alternatives, such as submerging entire data centers in the sea and using seawater for cooling, are already being actively implemented. However, these options are very expensive and complicated.
The energy consumption of data centers is a major problem: To operate their massive AI data centers, tech giants are now relying on old airplane turbines. While these provide quick power, they also create new environmental and supply problems: Because your AI asks so many dumb questions, data centers are now buying turbines from airplanes to supply enough energy
