LiquidStack has booked an order for 300 megawatts of coolant distribution unit capacity from a single US data center operator, one of the larger liquid-cooling commitments the company has disclosed publicly. The order covers its CDU-1MW units, deployed across multiple sites, though LiquidStack named the buyer only as a long-established operator building out AI-ready facilities across the country.
A CDU sits between a facility's chilled-water loop and the server racks it feeds. Direct-to-chip liquid cooling routes coolant straight to processor cold plates instead of relying on room air, and the CDU is the hardware that manages flow, pressure and temperature so the loop stays safe for the chips downstream. LiquidStack built the CDU-1MW for the density AI racks now demand, where per-rack power draw has outgrown what air handlers can remove economically.
"Orders of this size signal a clear inflection point for liquid cooling," said Joe Capes, CEO of LiquidStack. "Operators are committing to liquid cooling as core infrastructure for AI, and LiquidStack is uniquely positioned to support that transition at scale." The company did not disclose pricing or a delivery date for the order, saying only that its manufacturing and delivery capacity would support what it called aggressive build-out timelines across the customer's multiple sites.
Competing vendors are chasing the same market. Vertiv, Schneider Electric and other established cooling suppliers are also expanding CDU lines for AI-scale deployments, though comparable order sizes from those rivals were not available for this one. The company points instead to its own recent momentum: inclusion on Nvidia's recommended vendor list for CDUs, and an expanding manufacturing footprint in Carrollton, Texas, where it is based.
Its own history stretches back to 2012, when the company says it built one of the first high-density liquid cooling systems on the market. The current lineup spans several product families beyond the CDU-1MW, including immersion tanks marketed as DataTank, EdgeTank and CryptoTank, alongside modular systems called MicroModular and MacroModular, sold into hyperscale, colocation, neocloud, enterprise, edge and blockchain facilities. That range matters to a buyer weighing vendor risk: a supplier with deployment history across several facility types offers different assurance than one with a single reference customer, even before this new order ships.
For buyers weighing a shift to liquid cooling, an order this size is a data point on how quickly the supply side can scale to meet multi-site AI rollouts, even though this particular contract's price and delivery schedule remain undisclosed. Facilities engineers evaluating CDU vendors should still get firm commissioning dates and per-site capacity guarantees in writing, since a headline megawatt figure alone does not confirm how fast any one site gets built out.
Leave a Reply