The rapid growth of AI, cloud computing, and edge computing drives demand for data center capacity. Hyperscale data centers are expanding globally, leading to higher server densities and increasing thermal loads. Regulatory and societal pressure to reduce energy use and greenhouse gas emissions is also intensifying, forcing operators to seek more efficient infrastructure solutions.
Cooling systems are among the most energy-intensive components of a data center, typically accounting for 40%–60% of total energy consumption. Their performance significantly impacts the facility’s power usage, reliability of IT hardware, and operational costs. As power densities in server racks climb beyond 100 kW, traditional air-based cooling methods are reaching their performance and efficiency limits. As a result, there is strong market traction for liquid cooling technologies (Figure 1) in data centers designed for AI (also referred to as AI factories or high-performance computing).
Figure 1: Taxonomy of data center cooling technologies. CRAC/CRAH: Computer room air conditioner/handler. RDHX: Rear door heat exchangers.
Liquid cooling system configurations and key players
Cooling capacity, energy efficiency, water use, capital and operational expenditures (capex and opex), total cost of ownership, and climate conditions at the location influence the selection of cooling systems and their configurations. In a liquid cooling approach, the technology cooling system (TCS, also referred as the secondary cooling loop) is separated from the data center’s facility cooling system (FCS, also referred as the primary cooling loop) via a coolant distribution unit (CDU), which facilitates heat transfer through a heat exchanger. This separation provides greater flexibility, allowing for the use of high-purity heat transfer fluids (HTFs) and alternatives to water. The TCS configuration varies by the cooling mechanism and type of HTF used. In addition to the heat exchanger, the CDU typically includes pumps, control systems, and filtration units to ensure the TCS operates efficiently.
The primary function of the FCS is to reject heat to the atmosphere or, in some cases, enable heat recovery. The configuration of heat rejection units depends on climate conditions and the ambient temperature. Wherever possible, economizers (not included in Figure 2) can be integrated into either the TCS or FCS to leverage available cold air or water, thereby enhancing the overall energy efficiency of the cooling system.
Figure 2: Representations of liquid cooling configurations.
Figure 3: Key players in data center liquid cooling.
DTC: Unmet needs, innovations, and opportunities
Among liquid cooling technologies, DTC is the most mature and widely deployed solution, accounting for an estimated 60%–70% of the liquid cooling market. A key advantage of DTC systems is their ability to integrate with existing server racks with minimal infrastructure modification, making them particularly attractive for AI-centric retrofits and incremental upgrades in traditional air-cooled data centers.
However, as chip thermal design power (TDP) continues to rise, often exceeding 1,000 W per chip, traditional DTC architectures are approaching their performance limits. Higher TDPs demand higher coolant flow rates (generally >1 L/min per cold plate), which introduces significant mechanical stress on piping, manifolds, and seals. Higher-flow velocities also increase the risk of erosion, material fatigue, and, ultimately, fluid leakage. Other system components such as thermal interface materials (TIMs), critical for efficient heat transfer between the chip and cold plate, must evolve to accommodate increased heat flux, surface warping, and thermal cycling stresses.
Innovation in the DTC value chain is concentrated in four critical areas:
- First, next-generation TIMs must offer high out-of-plane thermal conductivity while remaining mechanically compliant under thermal cycling. Companies like Carbice and NovoLINC have developed nanomaterial-based TIMs tailored for data center applications.
- Second, to enhance heat transfer while minimizing pressure drop, cold plates are optimized using generative design algorithms and manufactured using additive manufacturing. Fabric8Labs, Alloy Enterprises, and Conflux Technologies have developed novel manufacturing processes to fabricate complex geometries.
- Third, researchers are exploring nanofluids, lower-viscosity HTFs, and dielectric alternatives to expand the operational envelope. For DTC-1P systems, replacing water with dielectric liquids is a key area to improve thermal transport properties without compromising material compatibility or increasing system pressure.
- Fourth, innovators are piloting DTC-2P, which has lower volume requirements and therefore less system pressure and lower risk of leaks. The major concern is the use of fluorinated HTFs, which are PFASs and have high global warming potential (GWP). PFAS-free and low-GWP HTFs are a huge opportunity for chemicals and materials developers.
Besides these technical challenges with DTC, there remain commercial and operational challenges such as high capex, high maintenance requirements, and end-of-life fluid management that currently hinder the adoption of liquid cooling technologies.
Immersion cooling: Unmet needs, innovations, and opportunities
Immersion cooling (IC) is an emerging technology growing its market share, currently accounting for 20%–30% of the liquid cooling market. Unlike DTC systems that require cold plates and fluid routing to individual components, immersion systems fully submerge servers in dielectric fluids, eliminating the need for discrete cold plates and TIMs. While the capex for IC is high compared to DTC, the opex savings due to low energy use and high cooling capacity bring down the total cost of ownership.
Despite its thermal efficiency advantages, IC faces significant barriers. The dielectric fluids used are often fluorinated organics (e.g., PFAS-based with high GWP) or synthetic hydrocarbons, many of which are costly, chemically persistent, and environmentally concerning. Data center system integration is a challenge: Existing servers and motherboards are not designed for submersion, requiring conformal coating or specially designed components. Additionally, heat exchangers and external CDU systems for immersion require precise fluid management and are often bulky, costly, and maintenance intensive.
Innovation in IC is occurring along two vectors to address unmet needs:
- First, fluid chemistry R&D efforts are underway to identify PFAS-free, low-GWP fluids that maintain high thermal stability and electrical insulation. Collaborative initiatives between petrochemical companies and IC-1P technology providers include Castrol and Submer, Shell and Asperitas, and Eneos and Green Revolution Cooling.
- The second area for innovation is mechanical system redesign. Startups and large corporate players are developing new immersion tank geometries, modular CDU integration, and server form factors to ease maintenance and improve deployment flexibility.
Despite its promise for high-density and edge environments, IC is not yet a drop-in solution. Its deployment requires purpose-built infrastructure, specialized training, and close coordination between server OEMs, fluid suppliers, and facility operators. Until industry-wide standards and environmental safeguards around dielectric fluids are in place, IC is likely to remain a specialized solution, best suited for experimental deployments or thermally extreme workloads.
Lux Take
Companies in the chemicals and materials sectors should engage with DTC players to capitalize on rapid adoption in the short term and focus on improving performance and reliability of TIMs, HTFs, and supporting components. However, innovators with a long-term focus should target IC as there are several unmet needs when it comes to high-density racks at lower power usage effectiveness and water usage effectiveness. A collaborative effort to develop compatible materials is necessary: Companies seizing the market opportunity in cooling systems should partner with chip developers, server OEMs and cooling system providers early in the design cycle. These partnerships will enable continuous product validation throughout development, streamlining innovation, and accelerating market adoption.
To discuss these trends further, reach out to schedule a meeting with the Lux Team.