Nidec adapting cooling strategy for Vera Rubin era
Thu, 19th Nov 2026 (Today)
Nidec Corporation is adapting its data centre cooling strategy as Nvidia's next-generation Vera Rubin computing platform drives demand for higher-capacity liquid cooling systems and changes how equipment suppliers engage with infrastructure operators.
Nidec is seeing a shift towards in-row cooling distribution units (CDUs), which can deliver substantially greater cooling capacity than traditional in-rack systems, alongside demand for modular solutions that allow data centre operators to scale capacity according to their requirements.
Deputy Head of AI & IT Division SPMS at Nidec Corporation, Nobuyasu Nozawa, said the transition is reshaping the market for thermal management equipment and shifting relationships between suppliers, server manufacturers and data centre operators.
The Vera Rubin platform, which follows Nvidia's Blackwell architecture, is increasing the need for cooling infrastructure capable of supporting increasingly demanding AI workloads.
This is creating opportunities for manufacturers to supply larger cooling systems directly to data centre operators rather than relying exclusively on server manufacturers as customers.
"There is a significant shift happening from in-rack CDU into in-row CDU," Nozawa said.
Historically, in-rack CDUs have been a primary form factor for liquid cooling systems, with equipment installed within server racks and supplied through server manufacturers.
However, the move towards higher-capacity in-row systems is changing that model.
Unlike in-rack equipment, which is designed to fit within a server rack, in-row CDUs can occupy a footprint comparable to or larger than the server racks they support. Their size and capacity mean they are typically installed directly in the data hall rather than incorporated into server infrastructure.
As a result, cooling suppliers are increasingly engaging with data centre operators and infrastructure teams responsible for designing and managing facilities.
Nidec is now speaking more frequently with data centre customers, while continuing to work with server manufacturers.
The shift represents a change in the commercial structure of the market, as well as an evolution in cooling technology.
For suppliers, engaging directly with data centre operators can introduce different purchasing requirements and opportunities.
Instead of supplying equipment through a server manufacturer, vendors must increasingly account for the cooling requirements of entire facilities and the infrastructure supporting their AI deployments.
However, not every operator requires the full capacity offered by an in-row CDU.
Nidec is also developing modular cooling configurations for customers seeking capacity between that of conventional in-rack systems and larger in-row units.
Nozawa said some customers required more than the approximately 250 kilowatts available from an individual in-rack CDU but did not need a system capable of delivering two megawatts.
For these operators, Nidec offers a stackable CDU configuration that combines multiple smaller units to achieve a required cooling capacity.
Four 250kW CDUs can be combined to provide one megawatt of capacity, while five units can deliver 1.25MW.
The approach allows customers to select cooling capacity in increments rather than committing to a larger system that could exceed their immediate requirements.
Nidec is among a relatively small number of companies offering this type of stackable CDU solution alongside larger in-row systems.
The modular approach could provide an alternative for operators balancing the demands of AI infrastructure with the need to match cooling investment to deployment requirements.
As AI computing platforms increase in power density, data centre operators must consider not only the performance of their servers but also the infrastructure needed to remove heat reliably and efficiently.
Liquid cooling is becoming more important as conventional air-cooling approaches face limitations in handling the thermal output of high-performance computing systems.
For manufacturers such as Nidec, these developments are creating demand for a broader range of cooling equipment, from systems integrated into individual racks to larger installations serving multiple racks.
The transition towards higher-capacity systems also means thermal management suppliers must adapt their products and customer engagement strategies as data centre designs evolve.
Nidec's strategy combines larger in-row CDUs with modular stackable configurations, allowing it to address different cooling requirements as operators deploy increasingly powerful AI infrastructure.
Nozawa expects these changing requirements to influence how cooling equipment is specified, purchased and installed as the industry moves towards the next generation of AI computing platforms.