Data-center liquid cooling in 2026 is a three-tier market, not a single leaderboard: server-OEM integrated loops, independent DLC specialists, and component-tier fabricators split the addressable revenue by load profile and loop topology rather than by overall market share [S1][S2].
AI training racks routinely exceeding 60 kW per rack, hot-water reuse mandates in EU colos, and density-driven HPC retrofits are the three measurable loads pulling demand; per-rack heat load, loop topology, and the supplier's certification chain (ISO 9001 / ISO 14001) are the three selection criteria that actually separate candidates [S2][S3].
Tier 1: Server-OEM Integrated Loops and DLC Specialists
Lenovo Neptune 6th Gen is positioned as a server-OEM-integrated loop covering three sub-modes, full-system warm water cooling, Neptune Core for component-level (CPU/GPU/memory) cold plates, and Neptune Air for liquid-assisted air-cooled retrofits, with the rear-door rack water heat exchanger completing the family; Lenovo cites over 12 years of field deployment and up to 40% energy savings versus traditional air cooling [S2].
CoolIT Systems, profiled in Data Center Dynamics' 20-year retrospective, sits in the second tier as a vendor-neutral DLC specialist; its architecture uses warm liquid rather than cold air to capture heat, allowing most CRAC units and chillers to be removed from operation once rack power exceeds roughly 20 kW, a documented efficiency inflection point for AI training clusters and HPC workloads [S1][S2].
Tier 2: Component-Tier Fabricators and CDUs
Koolance represents the third tier, an ISO 9001 + ISO 14001 dual-certified custom liquid cooling vendor, with cold plates, sub-assemblies, heat exchangers, chillers, pumps, and quick-connect fittings in catalog; the ALX-15KA is a 15 kW floor-standing ambient CDU running on 100-240 V AC at 700 W max consumption, and the VLX-2000 is a 2 kW (2000 W) 110 V recirculating chiller, with an optional rack-mount adapter tray for the ALX-3000 released on 2026-06-05 [S2].
ToneCooling (China), LiquidStack (Netherlands), Boyd Corporation (USA), and Schneider Electric (France) round out the systems-and-cold-plate field; ToneCooling publishes a 5-piece cold-plate prototype MOQ, copper-brazed thermal resistance of 0.04 degC/W at standard flow rates, and integrated support for both NVIDIA GB200 and Chinese GPU platforms (MetaX, Biren, Sunway, Hygon), with a stated 500,000+ units/year capacity [S3]. LiquidStack and ZutaCore cover two-phase immersion and dielectric waterless niches respectively, and Schneider supplies CDU and infrastructure at the systems tier [S3].
Spec Comparison: Material, Process, and Loop Topology
Cold-plate material selection, copper versus aluminum, is the first hard engineering trade-off, and it is not vendor-agnostic because brazing process windows differ by alloy; Winshare Thermal specifies liquid cold plates made from copper or aluminum with internal fluid channels sized to maximize convection between coolant and the mounting surface, while Mstirling (sister brand of Metalli) lists both vacuum brazing and controlled atmosphere brazing for brazed plate-fin heat exchangers and ultra-low temperature cooling systems [S1].
Loop topology drives fluid selection, which drives materials and leak-detection discipline; buyers should pin dielectric versus deionised-water choices, define allowable coolant conductivity, and require materials compatible with the chosen coolant, since dielectric fluids pull different elastomer and connector specs than deionised water loops [S2]. For industrial auxiliary equipment, the price gap is sharp: a 316/304 stainless-steel pneumatic hydraulic high-shear homogenizer is listed at US$1,500-4,000 per set by Guangzhou Hone Machinery, and a 24V/48V 1 kW DC R134A mini water-cooled chiller with evaporative heat exchanger is available from Jinan Amrta Air Conditioning at US$7.50 per set at single-set MOQ, while a single-piece DLC rack is a six-figure USD line item [S1].
Heat-Load Thresholds and Rack-Density Drivers
The 20 kW/rack threshold is the documented boundary above which air cooling is no longer sufficient and warm-water DLC becomes the economic winner; racks below 20 kW/rack are over-spec for DLC, and racks above 60 kW/rack, the AI training cluster regime, are the dominant demand pull in 2026 [S1][S3].
The NVIDIA GB200/MI300X era is documented at 60-120 kW/rack with PUE 1.05-1.15 under direct-to-chip liquid cooling, and next-gen AI racks in 2026+ target 120-200 kW+ with direct-to-chip plus immersion hybrid loops and sub-1.10 PUE, per ASHRAE-aligned guidance compiled by ToneCooling [S3]. Single-CDU figures like the ALX-15KA's 15 kW are useful reference points, but racks above roughly 30 kW typically need either warm-water direct loops (Neptune class) or stacked CDUs [S2].
Regional Demand, Standards, and Sourcing Cues
Middle East facilities face extreme ambient heat that forces advanced cooling, with government renewable-energy and sustainability focus amplifying liquid cooling's efficiency advantage in Saudi Arabia and the UAE; Latin America, led by Mexico under USMCA nearshoring and Chile's Santiago cloud build-out, is a secondary regional driver on the same hyperscale logic [S4][S6].
Operators in hot regions therefore evaluate chillers, dry coolers, hybrid systems, water treatment, and N+1 redundancy as integrated parts of the liquid cooling architecture, and modular prefabricated solutions are favored where they shorten site construction and reduce dependence on local specialist labor [S5]. OEM-integrated vendors inherit server-platform certifications (safety, EMC on the rack PDU) but typically do not advertise a standalone ISO 14001 on the cooling subsystem, a gap Koolance fills with dual ISO 9001 + ISO 14001 certification at the component tier [S2].
Decision Gate: Matching Vendor Tier to Load Profile
Buyers should match vendor tier to load profile: under 20 kW/rack, DLC is over-spec and rear-door heat exchangers or air-plus-RDHx retrofits are the right answer; 20-60 kW/rack is the warm-water DLC sweet spot for CoolIT-class specialists and Neptune-class OEM loops; above 60 kW/rack, cold-plate MOQ, hydraulic compatibility with the chosen CDU, and zero-leak helium-tested qualification dominate the spec gate, and component-tier vendors with low MOQ (ToneCooling at 5 pcs) and CFD-backed designs win on prototype speed [S1][S2][S3].
The selection signal to track next is the 2026-06-05 Koolance ALX-3000 rack-mount adapter tray release cadence, plus any OEM-tier warm-water loop announcements targeting the 120-200 kW/rack envelope, since those two moves will redefine the Tier 1 versus Tier 2 boundary before the 2026-12 hyperscale capex close. For broader thermal context, the parallel demand story in Data Center Cooling 2026: Liquid, WUE, and Hyperscale Spending Surge traces the capex side, while Wafer Fab Equipment Procurement Strategy covers the upstream semiconductor capacity feeding the GPU-density curve that drives all of this.
Spec-level background on the components involved: pressure transmitter, flow meter, and industrial valve.