The technology

Heat in. Cold out. No compressor.

Adsorption cooling is a century-old thermodynamic cycle whose moment has finally arrived - because liquid-cooled AI racks now produce waste heat at exactly the temperature it needs.

The cycle

Water is the refrigerant. Waste heat is the pump.

A sealed, near-vacuum vessel pairs a solid adsorbent with water. Water evaporates at low pressure - that evaporation is the cooling. The adsorbent soaks up the vapor; low-grade heat then drives the water back off, it condenses, and the cycle repeats. Two beds alternate so chilled water flows continuously. Zero-GWP refrigerant, essentially no moving parts, almost no maintenance.

1 · In
Warm return water
45-65°C from the rack's liquid loop feeds the regeneration side.
2 · Desorb
Bed regenerates
Heat drives water vapor off the saturated composite adsorbent.
3 · Evaporate
Cooling effect
On the other bed, water evaporates under vacuum and chills the secondary loop.
4 · Reject
Closed-loop dry rejection
Condenser heat leaves via dry coolers - no evaporation, no water lost.
The material

Why zeolite-calcium-chloride composites.

Plain zeolite holds a lot of water but needs ~150°C to regenerate - too hot for waste heat. Silica gel regenerates cool but holds less. A salt-in-matrix composite - calcium chloride layered into a porous zeolite/silica host - does both: far more water per gram and a low regeneration temperature that matches warm-water DLC.

Refrigerant
Water (R718, zero GWP, non-toxic)
Regeneration temperature
~55-75°C matches DLC return water
Water uptake vs plain zeolite-13X
~4.2× working swing (modeled)
Cooling COP (thermal)
~0.76 modeled vs ~0.54 plain 13X at 75°C
Compressor
None heat-driven, ~no moving parts
Water rejection
Closed-loop / dry evaporation-free

Composite COP and uptake figures above are drawn from peer-reviewed modeling and lab work, not yet a field-scale chiller - they are labeled modeled deliberately. Proving durability of the salt composite against leaching and corrosion under thousands of hydrothermal cycles is the core engineering we are doing.

Why now, not a decade ago: adsorption chillers need a heat source in the 50-95°C band. Air-cooled servers never delivered it. GB300-class liquid cooling does - NVIDIA warm-water DLC returns coolant around 45-65°C, and ~90% of rack heat is now captured to liquid. The waste stream and the driving requirement finally line up.

Honest engineering

What adsorption is - and isn't - good at.

The trade-off we accept

Adsorption COP is thermal (0.2-0.6), well below a vapor-compression chiller's electrical COP of 3-6. But it runs on free waste heat, not grid electricity - so the right metric is displaced compressor kilowatts and avoided evaporative water, not raw COP.

Where it wins

Hot and water-stressed regions where dry coolers lose margin on 35°C+ days and evaporative water is the real constraint - plus trim and peak duty. That's where we deploy first, not temperate climates where dry coolers already coast.

Footprint is real

Adsorbent beds are bulky and batch-cyclic - individual units are modest capacity. We manifold rack/row modules rather than pretend a single vessel cools a hall.

A complement, not a rip-out

We sit on the DLC secondary loop beside the CDU. You keep your liquid cooling; we monetize the heat it currently throws away.

Want the deeper technical brief?

We'll walk your thermal team through the cycle, the composite data, and a site-specific model.

AdsoluteZero assistantAsk about the technology or a pilot