Flex Partners with NVIDIA on 800V Power Rack to Address AI Infrastructure Bottlenecks
Event summary
- Flex and NVIDIA have jointly developed an 800 VDC Power Rack designed for NVIDIA’s Vera Rubin platform.
- The new power rack addresses limitations of existing infrastructure unable to support megawatt-scale AI workloads.
- The architecture increases available compute power per rack from approximately 125 kW to up to 880 kW when paired with future GPU platforms.
- Flex’s global manufacturing network, including North American capabilities, will support production and deployment.
- The product will be showcased at NVIDIA GTC 2026, March 16–19.
The big picture
The announcement highlights the escalating power demands of AI workloads and the need for specialized infrastructure to support them. Flex’s partnership with NVIDIA positions the company to capitalize on this trend, expanding its AI Infrastructure Platform and offering a solution that avoids costly retrofits for existing data centers. This move underscores the increasing complexity and specialization required within the data center power infrastructure market.
What we're watching
- Adoption Rate
- The speed at which data centers adopt 800 VDC architectures will dictate Flex’s revenue growth and influence NVIDIA’s Vera Rubin platform success. Early adopters will likely be hyperscalers and large enterprises with significant AI investments.
- Competitive Landscape
- Other power infrastructure providers will likely respond to this development, potentially introducing competing solutions or partnering with alternative GPU vendors, creating pricing pressure and impacting Flex’s margins.
- Supply Chain
- Flex’s ability to maintain a resilient and geographically diverse supply chain will be critical to meeting demand and avoiding disruptions, especially given ongoing geopolitical uncertainties.
Related topics