Nvidia and Schneider Electric announced a strategic collaboration to develop 800-volt direct current systems for next-generation AI data centers. The partnership marks a significant shift in how large-scale computing facilities deliver and stabilize electrical power, replacing the conventional 54 VDC standard with a megawatt-ready 800 VDC architecture.
The move responds to the escalating energy needs of AI-driven data centers referred to as AI factories, where racks now exceed the power requirements of a megawatt.
This collaboration aims to provide high-efficiency, high-density power distribution while addressing grid stability and operational cost concerns tied to synchronized GPU workloads.
Why the 800 VDC shift matters for AI infrastructure
Traditional data centers rely on low-voltage power distribution, which results in heat loss, heavy copper demand, and inefficient power conversion.
With Nvidia’s AI computing clusters now consuming hundreds of kilowatts per rack, these limits have become unsustainable.
Schneider Electric and Nvidia’s joint work establishes an 800 VDC ecosystem capable of end-to-end power delivery from the grid to the GPU.
According to Nvidia’s technical briefing, the company’s next-generation AI architectures, such as Rubin Ultra and Kyber GPU systems, will require power densities of over one megawatt per rack to run efficiently.
The proposed 800 VDC transition is essential not only for scalability but for safety, cost savings, and streamlined energy storage integration.
Did you know?
High-voltage direct current systems like 800 VDC can cut copper usage by more than 40% compared to standard 54 VDC networks, reducing both cost and thermal losses.
How Nvidia and Schneider plan to power next-generation GPUs
The Schneider-built 800 VDC sidecar power system acts as the intermediary between external power sources and the server racks. It converts incoming AC power into 800 VDC for direct rack distribution, enabling power delivery at scale while maintaining predictable, validated performance.
Each sidecar can power racks up to 1.2 MW, supporting advanced Nvidia GPU clusters optimized for AI training and simulation.
Schneider Electric’s Chief Technology Officer for Data Centers, Jim Simonelli, emphasized that 800 VDC marks the natural evolution toward megawatt-scale rack designs, reflecting the company’s system-level approach that merges efficiency, safety, and resiliency.
What sets the new 1.2 MW sidecar system apart
The new Schneider Electric sidecar introduces several high-performance features. It includes modular conversion shelves, live-swap capable energy storage, and intelligent metering for proactive power management.
These features enable maintenance without downtime while ensuring compliance with the highest safety standards, tested under arc-flash and fault-current scenarios.
In practical terms, the sidecar reduces material use, simplifies power-conversion chains, and minimizes heat dissipation, three significant constraints in the design of next-generation AI centers.
It leverages compact hardware integration based on Schneider’s “Life Is On” sustainability framework, ensuring a reduced footprint and a lower carbon lifecycle impact.
ALSO READ | Apple warns ATT may be disabled in EU amid probes
How 800 VDC tackles efficiency and volatility challenges
Nvidia’s 800 VDC research indicates a two-pronged design approach: high-voltage distribution and multi-timescale energy storage integration.
Power is delivered at 800 volts DC directly to compute racks, eliminating multiple AC-to-DC conversion steps and improving overall system efficiency beyond 96%.
The architecture also facilitates modular battery systems placed close to servers, enabling rapid response to stabilize voltage swings from synchronized AI workloads.
Such storage segmentation mitigates the “power volatility” caused by simultaneous GPU activity during AI model training.
Short-duration capacitors absorb millisecond fluctuations, while longer-duration batteries handle seconds-to-minutes fluctuations, protecting grid integrity and lowering peak power-sizing costs.
Where the AI power ecosystem expands next
The collaboration extends beyond Nvidia and Schneider. Over 20 major energy and semiconductor firms, including ABB, Eaton, Infineon, Hitachi Energy, Siemens, and Vertiv, are co-developing 800 VDC-compatible components.
This global alignment signals the emergence of an AI-oriented power ecosystem under the Open Compute Project to standardize connector interfaces and voltage ranges.
Nvidia has outlined a phased roadmap with OCP to bring large-scale AI factories online by 2027, using the 800 VDC standard as the backbone for megawatt GPU systems.
The broader goal is to achieve sustainable, global energy parity across AI workloads while reducing carbon output in an era of exponential compute demand.
The Schneider–Nvidia collaboration demonstrates how deep electrical innovation underpins the next generation of computing infrastructure.
As AI continues to redefine data center design, power becomes the new performance driver shaping the world’s digital future.
Comments (0)
Please sign in to leave a comment