The transition to 800VDC is no longer just a theoretical discussion. Across the data center industry, growing AI workloads and increasing rack densities are driving serious conversations about how power is delivered to next-generation compute infrastructure.

Today, most of those discussions are focused on the whitespace, and for good reason. Racks and GPUs are consuming unprecedented amounts of power and reshaping facility design priorities.

But while the industry’s attention remains fixed on powering compute, a broader question is beginning to emerge: what happens when the transition to 800VDC extends beyond the rack to cooling infrastructure?

As native DC power architectures mature, cooling systems may follow a four-phase adoption path, from optional rack-level deployment to native DC rack cooling, broader whitespace adoption, and ultimately facility-wide DC-powered infrastructure.

Exactly how quickly each phase develops remains uncertain. Yet the pace of adoption may depend less on end-user demand and more on whether cooling manufacturers and their OEM suppliers can build the component ecosystem necessary to support native 800VDC operations.

Phase one: Optional rack-level cooling

The first phase of the 800VDC transition is already underway. Current efforts are concentrated on the whitespace, where the primary goal is maximizing power delivery to IT equipment. This includes GPUs, racks, and the power distribution technologies that support them, including busbar, busway, power shelves, and remote power panels.

This focus reflects today's reality. AI infrastructure is pushing power densities higher, forcing operators and technology providers to rethink traditional electrical architectures. NVIDIA’s Rubin Ultra platform, expected in 2027, is projected to push rack densities as high as 600kW. Much of the current innovation around 800VDC is aimed at solving this delivery challenge first.

At this stage, however, native 800VDC cooling is not a requirement. In-rack power shelves convert power to 50VDC, so existing ORv3 cooling is still operational. Many operators will also continue using familiar AC-powered solutions. This decision may be driven more by customer preference than by technical necessity.

Phase two: Native DC cooling becomes standard at the rack

As AI deployments scale and 800VDC architectures mature, rack-level DC cooling shifts from optional to essential. In this second phase, with power shelves no longer required, high-density compute environments will operate on native 800VDC power.

The goal: simplify power distribution, reduce conversion stages, and align cooling infrastructure with the power system serving the IT equipment.

At this point, maintaining separate AC and DC power systems at the rack starts to add unnecessary complexity. Native DC cooling becomes the preferred approach for supporting the next generation of high-density IT deployments.

Reaching this phase, however, requires more than operator demand. Cooling manufacturers will need access to a growing ecosystem of OEM components specific to 800VDC environments. Without commercially available motors, drives, controls, and supporting electrical systems, widespread deployment of native DC cooling will lag behind market interest, not because operators are unwilling, but because the parts aren’t yet available.

Phase three: Whitespace cooling goes native DC

The third phase represents a more significant transformation: broader whitespace cooling infrastructure moves to 800VDC.

As 800VDC busway distributes native DC power throughout the data hall, AC power may no longer be readily available within the whitespace. This shift extends the DC conversation beyond rack-adjacent equipment and into larger cooling assets including pumps and larger CDUs, some of which are already being engineered at multi-megawatt thermal capacities to keep pace with rising rack loads.

Some operators will continue supporting AC-powered equipment through dedicated electrical pathways. But that approach introduces additional infrastructure requirements that must be weighed against the benefits of a unified DC architecture.

By this stage, supplier readiness becomes a deciding factor. Larger cooling assets serving the whitespace require a broader range of DC-compatible components, and the availability of mature OEM solutions determines how quickly operators can move beyond rack-level applications.

Phase four: Facility-wide DC cooling

The final phase extends 800VDC adoption arc, in which solid state transformers (SSTs) replace onsite rectifiers, touches virtually all cooling infrastructure within the data center. If AI-scale deployments continue driving native DC architectures, chillers, cooling towers, dry coolers, and other large mechanical systems come under the same pressure to operate within the facility’s DC power ecosystem. At that point, native DC becomes the default power environment for cooling equipment throughout the site.

That does not mean AC equipment disappears. Some facilities will incorporate low-voltage transformers that allow traditional equipment to continue operating where it makes sense. Even so, the long-term direction points toward greater integration between data center power architectures and thermal management systems.

Reaching this final phase represents not only an evolution in facility design, but also a significant transformation of the cooling supply chain. Every major subsystem, from compressors to pumps to controls, would require robust 800VDC offerings from OEM partners.

The ecosystem challenge

Whether the industry ultimately advances through all four phases will depend on far more than cooling equipment manufacturers alone. Cooling manufacturers rely on a large network of suppliers and OEM partners.

Pumps, compressors, fans, drives, controls, and instrumentation would all need market-ready solutions capable of supporting native 800VDC environments. Today, that component ecosystem remains largely immature, and the standards to guide it are still taking shape.

This is the real bottleneck. It isn’t cooling system design – the engineering path is well understood. It’s the availability of native DC pumps, motors, drives, controls, and instrumentation at the scale AI infrastructure demands.

The likely path forward starts with industry alignment. Organizations like the Open Compute Project (OCP), or major hyperscale operators may eventually establish system specifications for 800VDC cooling infrastructure. Component suppliers would then have a clear target for development, enabling cooling manufacturers to integrate compatible technologies into complete solutions.

Looking ahead

The industry's immediate focus should remain on powering IT equipment. That challenge alone will continue driving innovation across data center power architectures in the years ahead.

But as the transition to 800VDC advances, the next chapter appears to unfold outside the whitespace, in four distinct phases.

The question may not be whether cooling infrastructure can operate on 800VDC. It’s whether the supplier ecosystem can evolve quickly enough to support that future. As AI-driven power demands continue to reshape data center architecture, the organizations preparing for that transition today will be best positioned when the market moves beyond the rack.

Learn about Legrand's data center cooling technologies and how they support efficient, scalable thermal management for AI and high-performance computing environments here.