Leading CDU Manufacturers for AI Data Centers in 2026
The GPU gets the attention. The cold plate touches the chip. But between the server and the building sits another piece of equipment that is rapidly becoming critical to the AI data center: the CDU.
A Coolant Distribution Unit connects the technology cooling system inside the AI rack with the wider facility cooling infrastructure.
At its simplest, a liquid-to-liquid CDU performs several jobs simultaneously:
Heat Exchange + Pumping + Temperature Control + Pressure Control + Filtration + Monitoring
It also separates the facility water system from the cleaner, tightly controlled coolant loop serving expensive AI electronics.
As rack density moves from tens of kilowatts toward 100 kW, 200 kW and beyond, CDU capacity is moving just as rapidly.
The market now spans compact rack-mounted units around 100 kW all the way to equipment rated above 2 MW — and vendors are already designing architectures that scale toward 10 MW and beyond.
The CDU Is Becoming the Thermal Hub of the AI Rack
A typical direct-liquid-cooling loop looks like this:
Facility Water → CDU → Rack Manifold → Hose / QD → Cold Plate → Return → CDU
The CDU therefore sits at the boundary between two worlds.
On one side is the facility:
chillers, dry coolers, cooling towers and building water systems.
On the other side is the IT equipment:
GPU cold plates, CPU cold plates, manifolds, hoses and quick disconnects.
That makes the CDU much more than a pump cabinet.
The CDU is increasingly the control center of the liquid-cooling loop.
From 100 kW to Multi-Megawatt Systems
One of the clearest trends in 2026 is the widening range of CDU capacity.
Compact in-rack systems are useful where a single rack needs localized control or where liquid cooling is being introduced incrementally.
Large in-row systems can support several high-density AI racks from one centralized unit.
And hyperscale deployments are pushing CDU capacity well into the megawatt class.
DATAAD 2026 CDU Supplier Overview
| Manufacturer | Representative 2026 Capacity | Architecture | DATAAD View |
|---|---|---|---|
| Schneider Electric / Motivair | 105 kW – 2.5 MW per CDU | L2L / rack and facility-scale | Broad portfolio with architecture scalable beyond 10 MW |
| Vertiv | 70 kW – 2.3 MW | L2L and broader liquid-cooling ecosystem | Strong integration with complete critical data-center infrastructure |
| CoolIT Systems | Up to 2 MW CHx2000 | Row-based L2L | Long-established specialist in direct liquid cooling |
| Delta | Up to 2.4 MW | In-rack / In-row L2L and L2A | Strong link between AI power and thermal infrastructure |
| Cooler Master | 100/120 kW in-rack to 2.5 MW class | L2L and L2A | Rapid expansion from component cooling into hyperscale infrastructure |
| LG Electronics | 600 kW validated / 1.4 MW platform | L2L DTC infrastructure | Unique chip-to-CDU-to-chiller integration opportunity |
| Rittal | 150 kW in-rack / 1 MW in-row | L2L plus L2A options | Strong rack, OCP and modular infrastructure integration |
| nVent | 800+ kW CDU800 | L2L | Rack infrastructure combined with liquid-cooling capability |
This table is not a market-share ranking. Capacity ratings are based on representative products publicly listed by manufacturers and can depend on inlet conditions, temperature approach, flow rate and coolant specification.
1. Schneider Electric / Motivair — Scaling Toward the AI Factory
Motivair by Schneider Electric has one of the broadest publicly stated CDU capacity ranges among current suppliers.
Schneider announced its MCDU-70 in January 2026 with cooling capacity of up to 2.5 MW. The company says its individual CDU portfolio spans approximately 105 kW to 2.5 MW.
Perhaps more significant is the architecture around those units.
Schneider says centralized control can allow its CDU portfolio to scale to 10 MW and beyond, reflecting the direction in which gigawatt-scale AI factories are heading.
Motivair is also integrated into a wider Schneider portfolio covering chillers, rear-door systems, cold plates, controls, electrical infrastructure and data-center management.
Why DATAAD is watching:
The strategic advantage is not simply a large CDU.
It is the ability to integrate:
Power + CDU + Facility Cooling + Controls
inside one broader AI infrastructure architecture.
2. Vertiv — From 100 kW-Class Systems to 2.3 MW
Vertiv is another major supplier positioned across both the IT cooling layer and facility infrastructure.
The company's CoolChip CDU portfolio currently spans approximately 70 kW to 2.3 MW. Its CoolChip CDU 2300 is a liquid-to-liquid unit delivering 2.3 MW of rated cooling capacity.
Vertiv positions the 2.3 MW unit for deployment either in-row or in nearby mechanical space, allowing fewer CDUs to serve large AI clusters.
The larger reason Vertiv matters is ecosystem breadth.
Its data-center portfolio extends across thermal systems, power distribution, UPS infrastructure and high-density liquid cooling.
Why DATAAD is watching:
As AI infrastructure becomes increasingly integrated, companies capable of designing both cooling and electrical infrastructure gain a strategic advantage.
3. CoolIT Systems — A Specialist Built Around Direct Liquid Cooling
CoolIT Systems differs from diversified infrastructure companies because direct liquid cooling has been its core business for years.
Its CHx2000 row-based CDU provides up to 2 MW of cooling capacity. CoolIT says the system can support up to twelve 120 kW NVIDIA GB200 NVL72 racks under the operating assumptions used for the product.
The CHx2000 operates around a stated design flow of 1.2 LPM per kW and fits within a 750 × 1200 mm footprint.
CoolIT's portfolio extends from cold plates and server loops through CDUs and facility integration.
Why DATAAD is watching:
CoolIT represents the specialist model:
Cold Plate → Server Loop → CDU
rather than approaching liquid cooling primarily from HVAC or electrical infrastructure.
4. Delta — Power and Cooling Converge
Delta Electronics is particularly interesting because AI power infrastructure and liquid cooling are increasingly converging.
At COMPUTEX 2026, Delta presented a next-generation 2.4 MW liquid-to-liquid in-row CDU designed for high-density AI infrastructure. The system includes N+1 redundant pumps and is positioned alongside Delta's next-generation 800 VDC power architecture.
Delta also offers more compact CDU architectures. Earlier products included a 4U in-rack liquid-to-liquid unit above the 135 kW class, illustrating the company's ability to address both single-rack and multi-rack deployment models.
Why DATAAD is watching:
Delta may benefit from one of the most important changes in the AI infrastructure market:
high-density power delivery and high-density cooling increasingly have to be designed together.
5. Cooler Master — Moving Far Beyond PC Cooling
Many consumers know Cooler Master primarily as a PC cooling company.
That description is increasingly incomplete.
Its current data-center portfolio includes compact liquid-to-air systems, 100 kW and 120 kW in-rack liquid-to-liquid CDUs, and megawatt-class in-row platforms.
Cooler Master's current product catalog also lists 1.4 MW and 2.5 MW liquid-to-liquid CDU platforms. Its 2.5 MW design lists a maximum secondary flow rate of approximately 1,500 LPM and support for facility water temperatures across ASHRAE W2–W4 classes.
The company's portfolio also spans cold plates and primary/secondary liquid infrastructure.
Why DATAAD is watching:
Cooler Master illustrates how rapidly the AI boom is expanding the definition of a thermal-management company.
A business once associated primarily with PC cooling is now competing in megawatt-class data-center infrastructure.
6. LG Electronics — The Korean Chip-to-Chiller Strategy
LG Electronics may become one of the most interesting new competitors in this market.
At Data Center World 2026, LG displayed a 1.4 MW CDU as part of a broader AI data-center cooling portfolio that includes cold plates, CRAH systems and large chillers.
In July 2026, LG also announced that its 600 kW CDU received NVIDIA AI infrastructure validation after evaluation against more than 100 technical criteria.
This matters because LG is not approaching the market solely as a CDU supplier.
Its emerging strategy connects:
Cold Plate → CDU → Chiller → Facility HVAC
LG describes this broader approach as a Chip-to-Chiller thermal portfolio.
Why DATAAD is watching:
If data-center cooling continues moving toward deeper integration between the IT liquid loop and facility HVAC, LG's established chiller and HVAC capabilities could become a meaningful competitive advantage.
7. Rittal — Rack Infrastructure Meets Liquid Cooling
Rittal approaches the CDU market from another strategic direction: the rack itself.
Its current DLC lineup includes a 150 kW in-rack CDU and a 1 MW in-row CDU, as well as a roughly 70 kW liquid-to-air architecture.
The 150 kW version fits into a compact 4U rack form factor, while the larger in-row architecture is designed to serve clusters of high-density racks.
Rittal has also emphasized modularity and Open Rack V3/OCP design principles in developing its megawatt-class cooling infrastructure.
Why DATAAD is watching:
The future AI rack is becoming an integrated system containing:
Compute + Power + Manifold + CDU + Rack Structure
Rittal already operates directly in several of those layers.
8. nVent — RackChiller Moves Toward 800 kW+
nVent is another infrastructure company expanding aggressively into direct liquid cooling.
Its RackChiller CDU800 provides more than 800 kW of cooling capacity under stated operating conditions and includes redundant pumps, variable-speed drives, filtration, leak detection and remote management interfaces.
nVent also offers smaller in-rack CDU architectures, allowing the company to participate in both rack-level and larger centralized liquid-cooling deployments.
Why DATAAD is watching:
Like Rittal, nVent can combine rack and enclosure infrastructure with the liquid loop rather than treating the CDU as an isolated mechanical product.
CDU Capacity Alone Does Not Tell You Which Product Is Better
One of the biggest mistakes in comparing CDUs is looking only at the headline:
1 MW vs. 2 MW vs. 2.5 MW.
A CDU rated for 2 MW under one temperature condition may not deliver the same capacity under another.
The usable thermal capacity depends on variables such as:
- facility-water supply temperature,
- technology coolant supply temperature,
- temperature approach across the heat exchanger,
- flow rate,
- pressure drop,
- coolant composition,
- and system redundancy requirements.
This means buyers should compare operating points, not simply nameplate megawatts.
In-Rack or In-Row?
The second major decision is CDU location.
In-Rack CDU
A compact CDU installed within or directly beside a rack offers localized cooling control.
Advantages can include:
- modular deployment,
- easy addition to selected AI racks,
- short secondary coolant paths,
- and independent rack-level control.
The disadvantage is that the CDU consumes valuable rack space.
In-Row CDU
A larger CDU placed beside a group of racks can serve multiple AI systems.
This can reduce the number of CDUs required and move pumps and heat exchangers out of the individual compute rack.
It is increasingly attractive for large AI clusters.
Rittal's portfolio clearly illustrates the distinction, with approximately 150 kW in-rack and 1 MW in-row options.
Liquid-to-Liquid or Liquid-to-Air?
Most hyperscale AI discussions focus on Liquid-to-Liquid — L2L.
The CDU transfers heat from the technology cooling loop into a facility-water loop.
This is attractive for purpose-built AI data centers with appropriate water infrastructure.
But not every building has facility water available.
Liquid-to-Air — L2A systems transfer server heat from liquid back into facility air through radiators or integrated heat exchangers.
L2A can therefore provide a transition path for brownfield or retrofit sites.
Cooler Master, Rittal and other manufacturers now offer both forms within their portfolios.
What Buyers Should Compare
A serious CDU comparison should include much more than rated capacity.
1. Cooling capacity at the actual operating temperature
Ask for the performance curve at the facility-water and TCS temperatures planned for the project.
2. Maximum secondary flow
Future high-density racks can demand enormous coolant flow.
3. Available pressure
The pump must overcome pressure losses across manifolds, hoses, QDs and cold plates.
4. Pump redundancy
N+1 architectures are increasingly important in mission-critical AI infrastructure.
5. Heat-exchanger material
Material compatibility must match coolant chemistry and water-quality requirements.
6. Filtration and fluid monitoring
Particle control, conductivity and coolant chemistry become important as small cold-plate channels are introduced.
7. Leak detection
A CDU serves extremely expensive electronics.
Monitoring should therefore extend beyond the pump itself.
8. Controls and protocol support
Modbus, BACnet, SNMP, Redfish and other interfaces increasingly matter because CDUs are becoming part of the data-center control architecture.
9. Serviceability
Pumps, filters and sensors should ideally be serviceable without taking the entire AI cluster offline.
10. Scalability
A 300 kW requirement today may become 1 MW tomorrow.
Architecture matters as much as individual-unit capacity.
Do Not Forget the Manifold, Hose and QD
The CDU can deliver perfect temperature and pressure at its outlet.
That does not guarantee the GPU receives the required flow.
Between the CDU and cold plate sit:
Rack Manifold → Hose → Quick Disconnect → Server Manifold → Cold Plate
Every component creates pressure loss.
Every branch affects balancing.
Every seal introduces another reliability variable.
A properly engineered CDU therefore needs to be selected as part of the complete liquid loop.
Do not buy the CDU first and design the rack loop later.
The Market Is Moving From Products to Systems
The supplier landscape also tells us something about the direction of AI cooling.
Schneider/Motivair combines CDU and facility infrastructure.
Vertiv combines cooling and critical power.
Delta combines power electronics and thermal systems.
LG combines liquid cooling and HVAC.
Rittal and nVent combine rack infrastructure and cooling.
CoolIT approaches the problem from direct-liquid-cooling specialization.
Cooler Master is moving from component thermal engineering toward complete data-center systems.
The strategic direction is therefore clear:
The CDU is becoming less of a standalone machine and more of a component inside an integrated AI factory.
The Next CDU May Not Look Like Today's CDU
Today's mainstream CDU market is largely designed around single-phase water or water-glycol DLC.
But the cooling architecture may evolve again.
As DATAAD discussed in its special analysis on the potential 2028–2030 Two-Phase DLC transition, future systems may have to manage refrigerant, vapor return, condensation, working-fluid inventory and pressure control rather than simply circulating water.
That could blur the line between:
CDU + Refrigeration Unit + Heat-Rejection System
and create an entirely new generation of thermal infrastructure.
DATAAD View: Who Is Best Positioned?
There is no single winner because the answer depends on the project.
For a hyperscale facility seeking multi-megawatt integrated infrastructure, Schneider/Motivair and Vertiv are particularly significant.
For organizations seeking a long-established direct-liquid-cooling specialist, CoolIT deserves attention.
For projects where power and cooling integration is strategic, Delta is increasingly interesting.
For rack-centric modular deployment, Rittal and nVent offer a different value proposition.
For Asian AI infrastructure, Cooler Master's rapid expansion is notable.
And LG may be one of the most interesting companies to watch precisely because it can potentially connect the CDU to the much larger facility HVAC layer.
DATAAD Insight
Five years ago, many data-center operators rarely discussed CDUs.
Today, they are becoming fundamental AI infrastructure.
The reason is simple.
Once processors require liquid, somebody must control that liquid.
And as AI racks move from 100 kW toward several hundred kilowatts, the equipment controlling the liquid becomes progressively larger and more sophisticated.
The next competition will therefore not be decided only by who builds the largest CDU.
It will be decided by who can build the most reliable and scalable thermal system around it.
The cold plate cools the chip.
The CDU makes liquid cooling an infrastructure.
DATAAD 2026 Supplier Guide
This article is an editorial supplier overview, not a market-share ranking or purchasing recommendation. Listed cooling capacities are representative manufacturer specifications and can vary according to water temperatures, flow, pressure and operating conditions. Buyers should verify project-specific performance directly with each manufacturer.