A team of researchers at the University of Waterloo has made a breakthrough in developing a hybrid cooling system for AI data centers. The system uses a combination of deionized water and a specialized coolant to target hotspots on computer chips, potentially reducing the need for specialty coolant and improving efficiency. In a laboratory test, the researchers were able to cool the center of a simulated hotspot by about 17°C in just 15 seconds using a single drop of the coolant. The system works by submerging a laboratory heater in deionized water and placing a single drop of the coolant on its surface. When the drop boils, it rapidly cools the surrounding area, effectively targeting the hotspot. This approach is known as hybrid immersion cooling. The researchers, led by Tian-Yu Zhang, Arnav Banerjee, and Sushanta K. Mitra, believe that their system could be scaled up to work with larger systems, potentially reducing the amount of specialty coolant needed. Existing immersion cooling systems can carry away enough heat to warm a swimming pool, but they often struggle to target hotspots on computer chips. The Waterloo team's innovation addresses this issue by using a boiling drop of coolant to target the hotspots directly. This could lead to more efficient and cost-effective cooling systems for AI data centers, which are increasingly important for applications such as artificial intelligence, machine learning, and data analytics. The study was published in the Journal of Colloid and Interface Science.