Two-Phase Immersion Cooling

Two-phase immersion cooling exploits a phase change to move heat. Hardware sits in a sealed tank filled with a dielectric fluid engineered with a deliberately low boiling point — often around 50–60°C. When chip surfaces heat the surrounding fluid past that point, it boils directly on the silicon, and the rising vapour carries away large amounts of latent heat. At the top of the tank the vapour meets a condenser coil, gives up its heat to a facility water loop, condenses back to liquid, and rains back down over the hardware. The boiling itself does the transport work — no pump pushes fluid across the chips — making the primary heat-transfer loop elegantly passive.

Why the phase change is so powerful

The magic number is the latent heat of vaporization. Boiling a kilogram of fluid absorbs far more energy than merely warming it by a few degrees, so a two-phase system moves dramatically more heat per unit of fluid than single-phase circulation — and it does so while holding the chip surface near the fluid's boiling point almost regardless of load, a natural self-regulation that flattens hot spots. This is why the approach can sustain extreme power densities and post facility efficiency figures approaching a PUE of 1.02: almost no energy is spent on the cooling itself. On paper it is the endgame for the hottest next-generation silicon, mining ASICs included.

The catches that keep it rare in mining

Practice is harsher than theory, which is why two-phase remains far less common in Bitcoin mining than its single-phase sibling. The engineered fluorocarbon fluids that boil at just the right temperature are very expensive per litre. Because the working fluid constantly evaporates and condenses, the tank must be tightly sealed, and every service event — every time the lid opens to touch a hashboard — vents costly vapour that must be topped up, turning routine maintenance into a recurring expense. Many of these fluids also face environmental and regulatory scrutiny over atmospheric persistence and global-warming potential, with some chemistries being phased out by their own manufacturers. And the sealed condenser architecture simply makes board access harder than lifting a dripping unit out of an open oil bath: for an operator whose economics depend on fast, frequent hardware repair, serviceability is not a footnote — it is the business.

The heat-reuse angle

One genuine attraction for the heat-minded miner: the condenser delivers heat into a water loop at a usefully warm, remarkably stable temperature set by the fluid's boiling point. That consistency makes downstream waste-heat recovery — space heating, water pre-heating, greenhouse loops — easier to engineer than scavenging from variable hot air. The same virtue applies to hydro and single-phase systems, but two-phase's tight temperature regulation is the cleanest version of it.

Choosing between the immersion families

Fluid stewardship becomes an engineering discipline of its own in these systems: operators track vapor losses per service event, fit lids with vapor-recovery traps, schedule maintenance in batches to minimize openings, and monitor fluid chemistry for breakdown products that can form under hot-spot boiling. The fluid is effectively part of the capital equipment — treated, measured, and amortized like the tanks themselves rather than consumed casually like air.

The practical decision for most mining operators today still lands on single-phase immersion cooling: cheaper mineral-oil-class fluids, open tanks, forgiving maintenance, and proven results at scale, at the cost of pumps and somewhat lower ceiling density. Two-phase earns its complexity only where extreme density or ultra-low cooling overhead genuinely pays. For the broader context of tank-based cooling see immersion cooling, and see heat exchanger for how the captured heat finally leaves the building.