The race to build infrastructure capable of supporting artificial intelligence is beginning to redraw the boundaries between the technologies that power data centres and those that keep them cool.
That shift sits behind LITEON Technology’s decision to invest approximately $176 million in DCX Liquid Cooling Systems, a Warsaw-based specialist in cooling technology for AI and high-performance computing. Once the transaction closes, LITEON will hold about 25 per cent of the company.
On the surface, the deal brings together two complementary sets of technologies. DCX develops coolant distribution units, facility distribution units, cold plates and immersion cooling systems, while LITEON specialises in AI server power management, rack-level power delivery and 800VDC power architectures.
More significantly, it reflects a broader change in the design of AI infrastructure. As computing densities increase, power delivery and thermal management can no longer be treated as largely separate engineering disciplines. The performance of one increasingly determines the requirements placed on the other.
AI applications are driving higher rack power densities, increasing the importance of liquid cooling as operators seek ways to support more powerful computing environments. That creates a challenge that extends beyond removing heat from individual servers. Data centres must be designed around the interaction between computing hardware, electrical infrastructure and cooling capacity.
Power and cooling start to converge
LITEON and DCX intend to develop integrated power and cooling platforms for hyperscale data centres, while also collaborating on engineering, manufacturing and product development.
DCX’s portfolio illustrates the scale at which cooling infrastructure is now being developed. Its systems range from coolant distribution units rated between 600kW and 2.6MW to facility-level distribution units capable of handling up to 16MW, alongside cold plates, rack manifolds and modular immersion cooling systems.
The company, founded in 2019, says its products have been deployed with three of the world’s ten largest AI data centre operators and shipped to 75 countries.
For LITEON, the investment extends an existing AI infrastructure portfolio that already includes power management systems, battery backup units, racks, mechanical components and liquid cooling.
The strategic logic is therefore less about adding another individual product category than bringing electrical and thermal systems closer together. As AI hardware becomes more demanding, the traditional approach of optimising separate pieces of infrastructure may become increasingly difficult to sustain.
Tom Soong, chairman of LITEON, said data centre infrastructure had become a critical foundation for AI and that the investment would strengthen the company’s position across the AI data centre value chain.
Maciek Szadkowski, chief technology officer at DCX Liquid Cooling Systems, pointed to LITEON’s expertise in power electronics, manufacturing scale and relationships with hyperscale customers as central to the partnership.
AI changes the engineering equation
The implications go beyond the two companies involved. AI infrastructure is creating a different set of constraints for data centre designers and operators. More computing capacity does not simply require more servers. It creates corresponding demands for power delivery, backup systems, heat removal and the physical architecture surrounding the equipment.
That makes integration increasingly important. If power densities continue to rise, cooling cannot simply be added after decisions have been made about compute and electrical architecture. Equally, power systems have to be designed around the thermal characteristics and operating requirements of increasingly dense AI environments.
Liquid cooling therefore becomes more than a response to overheating. It forms part of the infrastructure architecture that determines how much computing capacity can realistically be deployed within a given facility.
Tomasz Buk, chief executive of DCX Liquid Cooling Systems, said the companies intend to offer operators an integrated solution designed for the next generation of AI computing platforms.
The significance of the LITEON investment lies in that direction of travel. AI is encouraging suppliers to think less in terms of isolated components and more in terms of complete infrastructure systems.
As data centres absorb increasingly powerful computing platforms, the dividing lines between power, cooling and compute are likely to become harder to maintain. The companies capable of coordinating those systems may ultimately be as important to the expansion of AI as the processors sitting inside the racks.
