Microsoft unveils advanced AI cooling which lowers heat, cuts energy use - and could lead to more powerful data centers (www.techradar.com)

Microsoft revealed lab tests of a microfluidics cooling system that routes coolant through tiny grooves etched directly into the back of silicon chips, rather than using conventional cold plates. In experiments the approach removed heat up to three times more effectively than cold plates and cut GPU temperature rise by about 65%. The design is augmented with AI-driven hotspot mapping to steer coolant precisely where chips run hottest. Microsoft says it has proven the concept and is moving next to reliability testing, working with Swiss startup Corintis and exploring integration with closed-loop water systems to curb consumption. For the AI/ML community this is significant because cooling limits are becoming a primary bottleneck as accelerators grow denser and more power‑hungry. Microfluidics could enable higher power-density boards, faster training/inference performance per rack, and materially lower energy and water costs for cooling—potentially delaying or avoiding thermal ceilings that could slow hardware scaling. The announcement also signals an industry race (Lenovo, Dell, Supermicro, Giga Computing and others) to innovate cooling, with implications for data center design, chip packaging and operational efficiency as models and hardware continue to scale.