Two-phase cooling uses the phase change from liquid to vapor to remove heat, leveraging the latent heat of vaporization. When liquid coolant boils, it absorbs enormous amounts of energy without changing temperature. This thermodynamic property makes two-phase cooling far more efficient per unit volume than single-phase liquid cooling where the fluid just heats up. Works above 1600W [[Thermal Design Power - TDP]]. At these power levels, single-phase liquid cooling can't remove heat fast enough. The coolant would need to flow at impractical velocities or temperatures would spike. Two-phase systems let the coolant boil directly on the hot surface. The vapor then condenses in a heat exchanger, releasing the captured heat. The condensed liquid recirculates back to the chip. Applied in both [[Direct-to-Chip Cooling]] and [[Immersion Cooling]]. For DTC, vapor chambers or heat pipes use two-phase. For immersion, the entire tank operates as a boiling bath where vapor rises and condenses on cold surfaces above. Physics determines the limits. Every cooling technology has a maximum heat flux it can handle. Two-phase cooling has the highest theoretical limit before you need exotic solutions like cryogenic cooling. Enables the next generation of AI accelerators. As [[Hyperscalers]] push toward 2000W+ chips, two-phase becomes mandatory. --- #deeptech #datacenters #firstprinciple