From "standard" to "rollout plan"

The 800VDC power architecture that NVIDIA is driving moved from a published standard to concrete deployment plans within just over a year of its Computex 2025 announcement. Once rack power exceeds 200kW, conventional 400V three-phase AC or 48/54VDC distribution hits the physical limits of copper busbar current capacity. NVIDIA has stated plainly that it will lead the industry's shift to 800VDC starting in 2027, with 1MW-plus IT racks in view.

The question is when, and who, will actually deploy it. Cross-referencing primary sources from power vendors and a major EMS (electronics manufacturing services) player shows that the rollout plan already has specific dates and sites attached.

Vertiv — products in H2 2026, engineering readiness in five months

Vertiv announced its alignment with NVIDIA's 800VDC roadmap in May 2025, and plans to ship its 800VDC power portfolio in the second half of 2026. The portfolio spans four categories: centralized rectifiers, DC busways, rack-level DC-DC converters, and DC-capable backup systems.

The speed is notable. Just five months after its May 2025 partnership announcement, Vertiv said in October of that year that it had reached engineering readiness on the 800VDC architecture in collaboration with NVIDIA. The product line's timing is designed to align with the 2027 rollout of NVIDIA's Rubin Ultra platform (Kyber racks).

  1. May 2025
    Vertiv announces strategic alignment with NVIDIA's 800VDC roadmap
  2. October 2025
    Vertiv reaches engineering readiness / Schneider Electric commits to 800VDC / Foxconn and NVIDIA announce implementation at the OCP Global Summit
  3. H1 2026
    Early 800VDC deployment proceeds at Foxconn's Kaohsiung K-1 AI data center in Taiwan
  4. June 2026
    Schneider Electric and Foxconn announce an additional strategic collaboration
  5. H2 2026
    Vertiv ships its 800VDC power portfolio
  6. 2027
    NVIDIA's Rubin Ultra-based Kyber rack (576 GPUs per rack) begins shipping

Foxconn — Kaohsiung K-1 as the lead deployment site

Foxconn (Hon Hai Technology Group), the world's largest EMS provider, announced on October 14, 2025 at the OCP Global Summit 2025 that it would implement the 800VDC architecture in AI factories in collaboration with NVIDIA. The first site designated for this rollout is Foxconn's "K-1" AI data center project in Kaohsiung, Taiwan, positioned as a demonstration site for the group's AI server, data center, and renewable energy integration capabilities.

Foxconn subsidiary Ingrasys Technology showcased a solution at OCP Global Summit 2025 that integrates the NVIDIA GB300 NVL72 platform with an in-row CDU (coolant distribution unit). The GB300 NVL72 packs 72 GPUs per rack at 142kW — itself an example of the power density of the "last transitional generation" before the 800VDC shift.

On June 15, 2026, Schneider Electric and Foxconn announced a further strategic collaboration for next-generation AI data centers. Combining Foxconn's compute platforms, AI rack integration, and global manufacturing capacity with Schneider's power, cooling, and energy management technology, the companies aim for an integrated, ready-to-deploy package emphasizing "speed, efficiency, and predictability" across regions. Foxconn Chairman Young Liu said the industry needs to reinvent how AI infrastructure is designed, built, and delivered.

Schneider Electric — supporting hyperscalers including Google and Meta

On October 13, 2025, Schneider Electric announced expanded support for 800VDC power infrastructure for NVIDIA's next-generation GPUs. Its 800VDC sidecar in development is designed to support major hyperscalers including Google and Meta, and includes a "Live Swap" feature that allows component replacement without powering down. Jim Simonelli, the company's CTO for data centers, framed the 800VDC transition as "a natural evolution" driven by rising compute density.

Supply chain breadth and deployment benefits

NVIDIA's 800VDC ecosystem includes major semiconductor makers such as Analog Devices, Infineon, Innoscience, MPS, Navitas, OnSemi, Renesas, ROHM, STMicroelectronics, and Texas Instruments. More than 50 MGX (NVIDIA's modular GPU reference design) partners are working on support for the new Vera Rubin NVL72 rack server generation.

NVIDIA's published figures for the deployment's benefits include up to 5% improvement in end-to-end efficiency, up to 70% reduction in maintenance costs, and up to 30% reduction in total cost of ownership (TCO) — driven mainly by fewer power-conversion stages and reduced conductor usage.

For procurement, power design, and investment decisions, 800VDC is no longer a "future concept." Two major power vendors, Vertiv and Schneider Electric, are bringing products to market within 2026, and Foxconn is already moving forward with early deployment at a real site. With Kyber racks set to ship in 2027, facility-side distribution design and equipment procurement lead times need to be factored in now.

Referenced FactCards