AI data centers are pushing cooling into a new category
As rack power densities climb, the challenge is no longer simply attaching a liquid-cooling system to individual servers; operators increasingly need to think about cooling as a facility-wide utility. That is where DCX Liquid Cooling Systems is positioning it’s latest technology and where it’s recent strategic investment from LITEON becomes significant.
In September, LITEON announced a strategic investment that will give it approximately 25% ownership of DCX in a transaction valued at around $176 million. The companies said they plan to combine DCX’s liquid-cooling expertise with LITEON’s capabilities in AI server power management, rack-level power delivery and 800VDC architectures, including the development of integrated power and cooling platforms for hyperscale data centers.
Cooling is becoming infrastructure
DCX’s latest focus is its Facility Distribution Unit, or FDU. Rather than deploying multiple smaller coolant distribution units around individual rows, the architecture moves liquid-cooling infrastructure toward the facility level. It’s latest 8MW-class FDU is specified at up to 8.15MW of heat-transfer capacity, with a 2°C approach temperature. DCX says the system can support 45°C warm-water operation, targeting high-density AI environments where traditional cooling architectures can become increasingly difficult to scale – the significance for operators is architectural.
A facility-scale CDU can consolidate pumps, heat exchangers and controls rather than replicating that equipment across numerous rows, that can simplify the cooling topology and potentially make large AI halls easier to manage, expand and service. Centralisation also raises the stakes – DCX’s latest discussion highlights coolant quality, monitoring and treatment as critical considerations when a single system is responsible for the thermal health of an entire hall.
The power and cooling link is getting closer
The LITEON investment points to another important development: power delivery and thermal management are increasingly being treated as connected infrastructure problems and AI clusters are forcing operators to plan electrical capacity, rack architecture and cooling capacity together. A high-density GPU deployment cannot simply be powered at scale and have cooling designed around it later. The combination of LITEON’s power expertise and DCX’s cooling portfolio (which includes CDUs, FDUs, cold plates and immersion systems) reflects that convergence. For data center executives, the broader question is therefore not just how much cooling capacity is available, but where that capacity sits in the architecture.
Why this matters to data center operators
This evolution reflects a wider shift in AI infrastructure; as compute density rises, cooling is becoming less of a supporting mechanical system and more of a core part of the data-centre architecture. The companies able to integrate cooling capacity, power delivery, controls and heat rejection at increasing scale will increasingly shape how AI facilities are designed. For operators planning the next generation of AI halls, DCX’s recent moves are another indication that thermal infrastructure is moving upstream, from the rack to the facility itself.
