NVIDIA Contributes Vera Rubin Architecture to OCP to Standardize Gigawatt-Scale AI Cloud Infrastructure
As modern artificial intelligence clusters rapidly transition from megawatt power footprints to gigawatt-scale AI factories, building and operating cloud infrastructure demands unprecedented levels of efficiency, modularity, and open standardization. To address these surging physical and computational bottlenecks, NVIDIA announced at the 2026 Open Compute Project (OCP) Global Summit in San Jose that it is contributing core rack and compute tray designs from its Vera Rubin platform to open standards via the NVIDIA MGX architecture.
Unifying Gigawatt Data Center Specifications
NVIDIA's contribution allows third-party server builders, original design manufacturers (ODMs), and hyperscale cloud providers to build upon standardized Vera Rubin NVL72 rack specifications. Designed to support massive agentic reasoning workloads and extreme memory throughput, the shared architecture provides modular reference designs capable of handling thermal requirements up to 250 kW per rack. Moving beyond traditional alternating current setups, the ecosystem is actively shifting toward 800-volt direct current (800V DC) power infrastructure, significantly cutting power loss across large server farms.
Broad Cloud and Hardware Ecosystem Collaboration
The OCP announcement represents a major collaborative push across the cloud and AI landscape:
- Google Cloud & NVIDIA: Expanding high-bandwidth fabric capabilities by announcing support for Falcon Transport on upcoming ConnectX-10 SuperNICs and BlueField data processing units (DPUs).
- Meta & NVIDIA: Enhancing open AI networking standards with innovations across the NVIDIA Spectrum-X Ethernet SuperNIC and Spectrum-6 switching platforms.
- Cisco & NVIDIA: Developing the Ultra-Dense Small Form Factor Pluggable (USFP) specification, establishing an open industry standard for 400Gb/s-per-lane optical networking fabrics.
- Open-Source Tooling: Launching open components for the NVIDIA DSX AI Factory software suite, giving cloud operators open-source tools for bare-metal cluster validation, automated provisioning, and unified rack management.
By democratizing high-density compute rack configurations and power delivery specifications, cloud hyperscalers and enterprise operators can deploy resilient, high-density AI clusters faster while avoiding proprietary vendor lock-in at the physical layer.