NVIDIA MGX

The Fastest, Most Flexible Path to AI Factories AI workloads demand more than traditional servers can deliver. NVIDIA MGX™ provides a modular reference architecture that enables OEMs, ODMs, and ecosystem partners to build accelerated systems faster. From single-node servers to rack-scale AI factories, NVIDIA MGX reduces engineering costs, shortens time to market, and delivers multi-generational compatibility for the entire ecosystem.

Explore what’s driving data center changes and a vision for a path forward, where power is distributed from the grid to the chip for high-power-density AI factories.

The NVIDIA Vera CPU is built for data movement and agentic reasoning, pairing seamlessly with NVIDIA GPUs for AI systems or operating independently to power analytics, cloud, orchestration, and storage workloads.

Go Inside MGX With MGX, OEM and ODM partners can build tailored solutions for different use cases while saving development resources and reducing time to market. The modular reference architecture allows for different configurations of GPUs, CPUs, and networking—including NVIDIA Grace™, NVIDIA Vera, x86, and other Arm® CPU servers—to accelerate diverse data center workloads.

Explore Technological Breakthroughs Modular Reference Design MGX offers over 100 combinations of modular designs, ranging from single-node servers to rack-scale solutions. Multi-Generational It ensures today's design investments are compatible with future GPU, CPU, DPU, and networking products. Open Ecosystem NVIDIA's MGX rack design and reference files for the Open Compute Project (OCP) enable ecosystem adoption, supporting both Arm and x86 servers. Accelerated Time to Market MGX helps partners deploy faster, lower risk, and build AI factories at scale. Standards Compatible MGX, supporting open standards such as PCIe, OCP, and Electronic Industries Alliance (EIA) rack specifications, integrates quickly into enterprise and cloud data centers. Full NVIDIA Software Stack MGX systems excel with the full NVIDIA software stack for AI, high-performance computing, and digital twins, allowing partners to offer diverse solutions.

NVIDIA RTX PRO™ Servers deliver breakthrough performance and energy efficiency for enterprise AI factories. With support for advanced NVIDIA networking like NVIDIA BlueField-3 DPUs and ConnectX®-8 SuperNICs , RTX PRO Servers enable seamless performance and scalability for agentic and industrial AI applications.

MGX Enables AI Infrastructure for AI Factories AI Factory AI factories are the new unit of compute—data centers designed to transform massive volumes of raw data into intelligence across the full AI lifecycle. Scaling AI performance requires a full-stack solution with CPUs, GPUs, networking, memory, and the entire data center ecosystem working together seamlessly. NVIDIA MGX provides the modular reference architecture that enables OEMs, ODMs, and ecosystem partners to efficiently deliver diverse solutions, from single node to rack scale. By standardizing building blocks across servers and racks, MGX reduces engineering effort, accelerates time to market, improves ROI, and ensures compatibility across generations—powering AI factories everywhere. NVIDIA 800 VDC Architecture The AI factories of the future will need to support megawatt (MW)-scale racks for maximum performance and efficiency, making power infrastructure the primary dictating factor for scale, location, and feasibility. This level of performance will require a fundamental architectural shift, not just incremental improvements. The NVIDIA 800 VDC architecture involves a high-voltage 800 VDC power distribution system with integrated, multi-time-scale energy storage—building the foundation for future computing needs.

Read the Latest on NVIDIA MGX Building the Modular Foundation for AI Factories With NVIDIA MGX With NVIDIA MGX, partners can use a building-block approach to design multiple systems, saving development costs and time to market. NVIDIA 800 VDC Architecture Will Power the Next Generation of AI Factories NVIDIA is leading the transition to 800 VDC data center infrastructure to support MW-scale IT racks and beyond, starting in 2027. Building the 800 VDC Ecosystem for Efficient, Scalable AI Factories Generative AI has transformed traditional data centers into AI factories, requiring a complete redesign centered on efficiency, scalability, and high energy demands.