Self-actuated thermomagnetic agitator for advanced immersion cooling in data centers

The continuous increase in power density of artificial intelligence (AI) chips and data center processors demands highly efficient thermal management solutions. While immersion cooling provides higher heat dissipation capability than conventional air cooling, its efficiency remains constrained by additional energy consumption required for mechanical pumping or condensing loads. Here, we introduce a self-actuated thermomagnetic (TM) agitator that harnesses waste heat from electronic devices to generate mechanical motion, thereby inducing localized fluid mixing. The agitator, composed of a magnetic material mounted on elastic cantilevers, undergoes periodic vertical oscillations near the Curie temperature due to temperature-dependent magnetic phase transitions, thereby enhancing convective heat transfer without any external power input. Theoretical analysis and structural optimization were conducted to determine the optimal operating conditions. Experimental results revealed that the TM agitator enhanced the convective heat transfer coefficient by up to 81% compared to natural convection cooling. The GPU chip demonstration showed that identical performance could be achieved with more than 30% less power consumption. These findings indicate that the integrating the TM agitator into existing immersion cooling platforms can augment cooling efficiency without additional power usage and shows a promising route for the thermal management of high-power density electronics and future data centers. • Gd magnet converts waste heat to mechanical work for fluid mixing. • Heat transfer coefficient reached up to 1.81 times that of natural convection. • A GPU cooling test confirms a 30% reduction in computational power consumption.