NVIDIA Contributes Blackwell Design to OCP, Boosting AI Innovation

In a move aimed at advancing open, efficient, and scalable data center technologies, NVIDIA has announced contributions from its NVIDIA Blackwell accelerated computing platform design, specifically the GB200 NVL72 system, to the Open Compute Project (OCP). This system, built on NVIDIA’s Blackwell architecture, includes innovations in rack architecture, liquid cooling, and compute tray mechanicals. These elements are designed to support higher compute density and increased networking bandwidth for AI infrastructure.

NVIDIA's History of Contributions to OCP

NVIDIA’s contributions to the OCP ecosystem are not new. The company has a long history of involvement, including the sharing of its NVIDIA HGX H100 baseboard design in previous hardware generations. These efforts aim to expand the range of offerings from computer manufacturers and increase the adoption of AI across the data center industry.

Expanded NVIDIA Spectrum-X Support

Alongside the hardware contributions, NVIDIA Spectrum-X Ethernet networking platform now aligns more closely with OCP standards. This enables organizations deploying AI factories to maximize the performance of their equipment while maintaining software consistency and preserving previous investments. A major component of Spectrum-X is the ConnectX-8 SuperNIC™, which offers advanced networking capabilities crucial for AI workloads.

ConnectX-8 SuperNIC: High-Speed Networking for AI

The ConnectX-8 SuperNIC is designed to support AI factories by providing ultra-fast networking with speeds of up to 800Gb/s. Its programmable packet processing engines are optimized for massive-scale AI workloads, enabling adaptive routing and telemetry-based congestion control. This ensures efficient data flow, even at high compute scales, making it a critical component for AI-driven data centers. Additionally, the ConnectX-8 supports OCP’s Switch Abstraction Interface (SAI) and Software for Open Networking in the Cloud (SONiC) standards, further enhancing its compatibility and flexibility for open infrastructure deployments. ConnectX-8 SuperNICs for OCP 3.0 will be available starting next year, equipping organizations to build scalable, flexible networks optimized for AI.

Jensen Huang on the Future of AI Factories

NVIDIA’s founder and CEO, Jensen Huang, emphasized the importance of collaboration with OCP in shaping the future of data centers: “By advancing open standards, we’re helping organizations worldwide take advantage of the full potential of accelerated computing and create the AI factories of the future.”

Accelerated Computing for AI-Powered Data Centers

The NVIDIA GB200 NVL72 system design is built on NVIDIA MGX™ modular architecture, connecting 36 NVIDIA Grace™ CPUs and 72 NVIDIA Blackwell GPUs in a powerful, liquid-cooled, rack-scale system. This design allows the system to function as a single, massive GPU, offering up to 30x faster real-time performance for large language model inference compared to NVIDIA’s previous generation GPUs.

Partnerships and Industry Collaboration

As NVIDIA collaborates with more than 40 electronics manufacturers globally, the company is driving the development of AI factories—highly efficient data centers optimized for AI workloads. Partners like Meta are already building on top of the NVIDIA Blackwell platform, with plans to contribute the Catalina AI rack architecture to OCP. This provides flexibility for computer makers to design systems that meet growing performance and energy efficiency requirements.

What This Means for the Future of AI Infrastructure

NVIDIA’s contributions to the Open Compute Project signal a major advancement in the development of scalable, high-performance AI infrastructure. By opening up key elements of its Blackwell platform, including the GB200 NVL72 system, NVIDIA is accelerating the adoption of AI across the data center industry. The addition of ConnectX-8 SuperNICs further enhances networking capabilities, offering ultra-fast data transfer optimized for large-scale AI workloads.

These open standards will not only drive innovation in AI hardware but also help ensure that organizations worldwide can benefit from AI advancements without being locked into proprietary systems. As the demand for AI infrastructure grows, NVIDIA’s partnership with OCP will likely set the stage for the next generation of AI factories, enabling a more open, flexible, and efficient future for AI-driven computing.