The major liquid cooling data center companies include Vertiv, Schneider Electric, CoolIT Systems, Asetek, GRC (Green Revolution Cooling), Submer, LiquidStack, Iceotope, Boyd, Motivair, and JetCool. Each targets a distinct part of the market, from direct-to-chip cold plates on individual servers to full-facility immersion tanks housing entire racks.
GPU and AI accelerator density is the reason this market expanded so quickly. As hyperscaler infrastructure spend shifted toward dense AI clusters, the thermal output per rack climbed well beyond what traditional air-based systems can sustain. A single rack of NVIDIA H100 or B200 GPUs can exceed 100 kW; air cooling tops out at roughly 20-30 kW per rack in most production deployments. That gap is why liquid cooling went from a niche option to a procurement requirement at scale.
Direct-to-Chip Liquid Cooling Vendors
Direct-to-chip (DTC) cooling, also called cold plate cooling, routes chilled water or coolant directly to a metal plate mounted on the processor. Heat transfers conductively from the chip to the plate, then away through facility cooling infrastructure. Most deployments use DTC alongside residual air cooling for storage, drives, and memory, making it compatible with existing data center builds without a full facility redesign.
CoolIT Systems
CoolIT Systems is a Calgary-based specialist that has been building DTC solutions since the mid-2000s. Their rack-level coolant distribution units (CDUs) are integrated into OEM server designs from major manufacturers, meaning you often encounter CoolIT technology inside branded servers without seeing the company name on the front panel. Their focus is high-volume OEM supply and fluid management at the rack level. You can review their product line at coolitsystems.com.
Asetek
Asetek, headquartered in Denmark, holds foundational patents in pump-in-loop liquid cooling and supplies both consumer and data center markets. In enterprise deployments, their rack liquid cooling (RLC) systems connect individual server cold plates to a rear-door heat exchanger or a centralized CDU. They have long-standing OEM agreements with server vendors across North America, Europe, and Asia, which gives them wide installed-base reach without always appearing prominently in end-user buying decisions.
JetCool
JetCool is a Massachusetts-based company focused specifically on high-heat-flux scenarios. Their microconvective cold plate technology uses targeted jet impingement to move heat more aggressively than standard conduction-based plates. The design is particularly relevant for AI accelerator workloads where heat concentration on a small die area creates localized hot spots that a conventional flat cold plate cannot address uniformly. JetCool entered the data center market primarily through GPU and custom ASIC cooling applications.
Motivair
Motivair builds rack-level and facility-level DTC infrastructure, including rear-door heat exchangers and CDU systems designed for high-density AI compute clusters. Their ChilledDoor product attaches to standard rack enclosures and captures heat in the air stream before it enters the facility cooling loop, which suits operators who want higher density without converting to full immersion. They also offer turnkey liquid-cooled rack systems aimed at HPC and AI deployments.
Immersion Cooling Specialists
Immersion cooling submerges servers, or components, directly into a thermally conductive but electrically non-conductive fluid. There are two main variants: single-phase immersion, where the fluid stays liquid throughout, and two-phase immersion, where the fluid boils at low temperatures and vapor condenses on a heat exchanger above the tank. Both approaches achieve thermal densities far beyond DTC and are increasingly specified for AI training clusters where rack density is the primary constraint.
Before committing to immersion, you should understand how it differs from DTC at the operational level. The traditional cooling methods used in most existing facilities are not simply swapped out; immersion requires either a purpose-built facility or significant retrofit work, including specialized tank infrastructure, fluid management, server compatibility verification, and maintenance procedures that differ considerably from air-cooled environments.
GRC (Green Revolution Cooling)
Green Revolution Cooling, commonly referred to as GRC, is an Austin, Texas company and one of the longer-standing names in commercial single-phase immersion cooling. Their CarnotJet system uses a mineral-oil-based dielectric fluid and supports standard off-the-shelf servers without hardware modification, which reduces the barrier to adoption compared to some two-phase alternatives. GRC has deployments spanning HPC, enterprise, and cryptocurrency mining environments. You can find technical documentation at grcooling.com.
Submer
Submer is a Barcelona-based company focused on single-phase immersion using their proprietary SmartPodX tanks and SubmerGreen fluid. Their architecture is designed with modularity in mind, scaling from single-rack pods to multi-megawatt deployments. Submer has positioned itself strongly in the European market and has attracted attention from colocation providers looking to offer liquid-cooled capacity without full facility rebuilds. Their fluid is biodegradable and non-toxic, which matters for permitting and waste management at scale.
LiquidStack
LiquidStack, which was spun out of Bitfury’s hardware group, operates in two-phase immersion cooling. Their systems use a low-boiling-point fluid that vaporizes on contact with hot components, then condenses on a cooled surface and returns as liquid. Two-phase systems typically achieve higher heat transfer coefficients than single-phase and can handle extreme rack densities. LiquidStack has worked with semiconductor companies and hyperscale operators, and their technology has been evaluated for AI cluster applications where single-phase capacity is insufficient.
Iceotope
Iceotope, a UK-based vendor, takes a different architectural approach with what they call precision immersion, where coolant is circulated at the chassis level rather than submerging an entire open tank. Individual server modules sit in sealed chassis filled with dielectric fluid, and heat transfers to facility water at the rack. This approach preserves a more familiar server form factor, reduces fluid volume per deployment, and makes individual server access cleaner than an open tank. They target both hyperscale and edge computing applications where physical constraints matter.
Full-Stack Infrastructure Vendors with Liquid Cooling Portfolios
Several large infrastructure companies have built or acquired liquid cooling capabilities as part of broader data center power and thermal product lines. These vendors compete differently from the specialists above; they offer liquid cooling as one component of a complete solution that also covers power distribution, monitoring, containment, and facility management.
Vertiv
Vertiv is a Columbus, Ohio company and one of the largest data center infrastructure vendors globally. Their liquid cooling portfolio includes rear-door heat exchangers, CDUs, and purpose-built liquid-cooled rack systems. Vertiv’s advantage is integration: their thermal products connect directly to their power management and monitoring platforms, which matters for operators who want a single vendor for power, cooling, and management infrastructure. They are widely specified by colocation providers and enterprise buyers who prioritise support continuity.
Schneider Electric
Schneider Electric, through its EcoStruxure Data Center and APC product lines, offers direct liquid cooling solutions including CDUs, in-row cooling units with liquid connections, and rack-integrated liquid cooling systems. Schneider’s scale means their products are evaluated in nearly every large infrastructure RFP. For operators already running Schneider power and monitoring infrastructure, their liquid cooling products integrate natively into existing management software, which reduces integration work during deployment.
Boyd Corporation
Boyd is a thermal management and fluid handling company supplying cold plate assemblies, heat exchangers, and fluid management components. They operate primarily as a manufacturing partner to OEMs and system integrators rather than as a direct end-user vendor. If you are evaluating a custom liquid cooling build for a dense AI cluster, Boyd components are likely in the supply chain whether or not the name appears in the specification.
How NVIDIA Reference Designs Are Shaping Vendor Selection
One factor that does not appear in most vendor comparison lists is the role of GPU manufacturer reference designs. NVIDIA’s specifications for the H100, H200, and Blackwell (B200, GB200) GPU systems include liquid cooling recommendations that effectively pre-qualify certain vendor approaches. OEM server builders follow NVIDIA’s thermal guidance closely, so vendors whose cold plate and CDU specifications align with the reference architecture have an easier path into high-volume AI cluster deployments.
This concentrates purchasing decisions at the OEM level. A hyperscaler buying tens of thousands of GPU nodes largely buys whatever thermal solution the server OEM validated against the GPU vendor’s specification. Direct vendor selection from the list above tends to occur in custom HPC builds, colocation facility design, and edge deployments where standard OEM form factors do not apply.
Frequently Asked Questions
Which companies make data center liquid cooling?
The main liquid cooling data center companies are Vertiv, Schneider Electric, CoolIT Systems, Asetek, Boyd, Motivair, JetCool, GRC, Submer, LiquidStack, and Iceotope. Some specialise in direct-to-chip cold plates; others focus on full-tank immersion. Large infrastructure vendors like Vertiv and Schneider offer liquid cooling as part of broader facility solutions.
Who are the leading immersion cooling companies?
The most prominent immersion cooling specialists are GRC (Green Revolution Cooling), Submer, LiquidStack, and Iceotope. GRC and Submer focus on single-phase systems using dielectric fluids at ambient pressure. LiquidStack uses two-phase technology where fluid boils at low temperatures. Iceotope offers sealed precision immersion chassis rather than open tanks, which suits deployments where server access frequency is high.
What is the difference between direct-to-chip and immersion cooling vendors?
Direct-to-chip vendors, including CoolIT, Asetek, and JetCool, supply cold plates that mount directly on processors and connect to facility water via a coolant distribution unit. Immersion vendors submerge entire servers in dielectric fluid. DTC is easier to retrofit into existing facilities and handles most AI workloads. Immersion supports higher rack densities but requires purpose-built infrastructure and modified server configurations.
Why is liquid cooling demand growing for data centers?
AI accelerator workloads produce heat densities that air cooling cannot handle at scale. A rack of modern GPU servers can exceed 100 kW of thermal output; typical air-cooled infrastructure manages 20-30 kW per rack. As hyperscalers and colocation providers deploy denser AI clusters, liquid cooling moves from optional to required. Processor power consumption continues to rise with each GPU generation, making the demand trajectory predictable.
How do you choose a liquid cooling vendor for a data center?
Start by matching the technology to your rack density and facility constraints. DTC works with existing infrastructure at lower upfront cost; immersion supports higher density but needs facility changes. Check whether the vendor’s system is validated by your server OEM for your specific GPU model. Evaluate support coverage, fluid management requirements, and whether the vendor integrates with your existing power and monitoring platform before shortlisting.
