Autumn production shipments for Vera Rubin

The Nvidia Vera Rubin platform will ship in production volumes in the autumn.

Vera Rubin delivers  five purpose-built racks operating as one AI supercomputer for agentic workloads.

The platform unifies  Vera Rubin NVL72 systems,  Vera CPU,  Groq 3 LPX,  Vera BlueField-4 STX storage and Nvidia Spectrum-6 SPX Ethernet racks into a fully integrated system.

Vera Rubin delivers 10x agent throughput at scale compared with the previous-generation NVIDIA Grace Blackwell platform.

“Agentic AI is a new kind of workload. One prompt can launch a thousand-step journey of reasoning, retrieval, tool use and response generation,” says Jensen Huang, “Vera Rubin was built for this moment — an AI factory engine that delivers intelligence at scale, with the performance, efficiency and security needed to power the next industrial revolution.”

Vera Rubin marks the third generation of Nvidia MGX rack-scale systems.

It is currently being built into systems by:  Dell, HPE, Lenovo, Supermicro, AIC, Aivres, ASRock Rack, ASUS, Cloudian, Compal, DDN, Everpure, Foxconn, GIGABYTE, Hitachi Vantara, Hyve Solutions, IBM, Inventec, MinIO, MiTAC Computing, MSI, NetApp, Nutanix, Pegatron, Quanta Cloud Technology (QCT), VAST Data, WEKA, Wistron and Wiwynn.

The platform introduces m Spectrum-X Ethernet Photonics – co-packaged-optics (CPO)-based switches with 200Gb/s SerDes.

By simplifying design and freeing more power for compute, NVIDIA co-packaged optics networking provides the foundational fabric for million-GPU AI factories, with CoreWeave, Lambda and Oracle Cloud Infrastructure among the first ecosystem partners and adopters.

 Vera Rubin  also integrates Nvidia BlueField-4 DPUs, featuring software-defined networking at speeds of up to 800Gb/s and built-in multi-tenant isolation. With the  BlueField-4 Advanced Secure Trusted Resource Architecture, customers can simplify network operations, improve tenant isolation and gain greater control across million GPU AI clusters.

The Vera Rubin platform was designed with full-stack Nvidia Confidential Computing for a trusted execution environment at rack scale. Vera Rubin NVL72 combines Vera CPUs, Rubin GPUs, NVIDIA NVLink networking and security features into a unified platform, encrypting data across high-speed interconnects. This provides hardware-level attestation to ensure the system is tamper-proof.

Cloud providers CoreWeave, Firmus, GMI Cloud, IBM Cloud, IREN, Lambda, Microsoft Azure, Nebius, Nscale, SpaceXAI and Vultr are adopting NVIDIA Confidential Computing.

Delivering this level of protection at POD scale also requires a programmable software layer capable of enforcing, orchestrating and adapting security policies across the entire system. The DOCA software platform delivers advanced security across every Vera Rubin platform rack and layer of the AI factory — protecting data, agents, context memory and AI inference through capabilities enforced directly in BlueField-4 silicon.

DOCA enables multi-tenant network isolation, zero-trust policy enforcement, runtime threat detection and end-to-end encryption at speeds of up to 800Gb/s, all without taxing host CPU resources, so enterprises can scale AI factories with confidence.

Accelerating the Buildout of AI Factories
The NVIDIA DSX platform provides the complete design and operational foundation for Vera Rubin AI factories — unifying reference designs, simulation, infrastructure software, facilities and ecosystem technologies to help build and operate energy-efficient AI factories optimized for lowest token cost.

Built for the Vera Rubin POD architecture, DSX aligns every layer of the stack — from silicon and systems to lifecycle management and multi-tenant operations — dramatically accelerating deployment and setting a new bar for operational reliability and resiliency at scale.

Dell Technologies, HPE, Lenovo and Supermicro together with ASUS, Foxconn, GIGABYTE, Pegatron, Quanta Cloud Technology (QCT), Wistron and Wiwynn are adopting NVIDIA DSX to accelerate AI factory ramp with Vera Rubin.