Vertiv Aligns Proactively with Nvidia’s AI Roadmap Through 800V HVDC Launch for Next-Gen Data Center Power
Vertiv Holdings (NYSE: VRT) makes the systems that power data centers. Its core business is delivering the high-density electrical infrastructure—power supplies, distribution systems, and backup architectures—that keep servers running continuously and efficiently. As data centers evolve to support AI workloads requiring far more electricity per rack, Vertiv’s role is expanding from traditional UPS systems to advanced high-voltage DC power solutions. The company isn’t just supporting the cloud—it’s enabling the power layer beneath AI clusters, hyperscale campuses, and the next generation of compute infrastructure.
Vertiv has announced a major step forward in AI data center power infrastructure: the planned launch of a full-stack 800V direct current (DC) power supply system, scheduled for the second half of 2026. The system will include centralized rectifiers, high-efficiency DC bus ducts, rack-level DC-DC converters, and DC backup systems. The release is timed to precede the deployment of Nvidia’s next-generation Kyber and Rubin Ultra platforms—signaling deep strategic alignment between Vertiv and the world's most important AI compute provider.
This development not only reinforces Vertiv’s role in the AI infrastructure ecosystem, but also validates a broader industry trend: the shift from traditional AC uninterruptible power supplies (UPS) to high-voltage DC (HVDC) systems that offer better efficiency, scalability, and cost advantages in AI-optimized data centers.
Comparison of Traditional AC UPS vs 800V HVDC Power Architecture
Table 1 shows the technical and economic differences between legacy AC UPS power infrastructure and the emerging 800V HVDC systems being adopted in AI-focused data centers. As server rack power exceeds 300kW, HVDC architectures are gaining favor due to efficiency and integration advantages.
Alignment with Nvidia’s Roadmap
Vertiv’s announcement was not vague or opportunistic. The company specifically stated its new 800VDC line would precede the rollout of Nvidia’s Kyber and Rubin Ultra platforms. These next-gen architectures are expected to push the limits of AI training and inference at scale—powering entire racks with GPU cabinets likely exceeding 40–60 kW per chassis and over 300 kW per rack. This will drive a full-stack redesign of data center power and cooling systems.
In launching ahead of this timeline, Vertiv ensures that customers building around Nvidia’s roadmap can design their physical infrastructure in parallel. This also supports Nvidia’s broader push toward cabinet-scale computing and prefabricated AI data centers.
AI Rack Power Evolution and Infrastructure Implications
Table 2 shows how AI server rack power density has increased over time and the resulting infrastructure changes needed to support these advances. Vertiv’s timing reflects a proactive move to address the next leap in energy delivery and thermal management.