NVIDIA announced on May 31, 2026, that the NVIDIA Vera Rubin platform continues to rise into full production to power agentic AI factories around the world.
It is well to be noted that the leading server manufacturers and global supply chain giants from Taiwan are deploying Vera Rubin-based systems at scale – fuelling AI labs and cloud providers as well as hyperscalers so as to build intelligence for tomorrow.
The biggest POD-scale platform of NVIDIA five purpose-built racks working as one giant AI supercomputer powers agentic workloads for Vera Rubin. The platform integrates NVIDIA Vera Rubin NVL72 systems, NVIDIA Groq 3 LPX, NVIDIA Vera BlueField-4 STX storage, NVIDIA Vera CPU, and NVIDIA Spectrum™-6 SPX Ethernet racks into one, seamlessly integrated system. Vera Rubin generates 10x agent throughput at scale versus the prior-generation NVIDIA Grace Blackwell platform.
According to the founder and CEO of NVIDIA, Jensen Huang, “agentic AI is a new kind of workload. One prompt can launch a thousand-step journey of reasoning, retrieval, tool use and response generation. 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 Bridge
Vera Rubin is the third generation of the NVIDIA MGX™ rack-scale systems. NVIDIA has a supply chain ecosystem that includes 150 partners in Taiwan. They are ramping up Vera Rubin across more than 350 factories in 30 countries utilising a validated, freely available open-source MGX design.
Vera Rubin is in full-scale production with top system builders and infrastructure software, along with storage partners. This includes the likes of Dell Technologies, Lenovo, Supermicro, and HPE, as well as 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, and Wistron as well as Wiwynn.
Constructing the Fabric for Million-GPU AI Factories
Apparently, the Vera Rubin platform employs NVIDIA Spectrum-X™ Ethernet Photonics, which is the first co-packaged-optics – CPO-based switches in the world with 200Gb/s SerDes, currently in production, so as to power scale-out and scale-across AI factory installations.
Spectrum-X Ethernet Photonics, a new breed of switching technology based on CPO, provides 5x higher power efficiency and 5x longer AI uptime as well as 1.3x faster time to installation compared with legacy networks with traditional transceivers.
Interestingly, NVIDIA co-packaged optics networking streamlines design and enables greater power for computation, providing the foundation fabric for million-GPU AI factories. Lambda and CoreWeave, as well as Oracle Cloud Infrastructure, are among the initial ecosystem collaborators as well as adopters.
The NVIDIA Vera Rubin platform additionally includes NVIDIA BlueField®-4 DPUs, offering software-defined networking at almost 800Gb/s and integrated multi-tenant isolation. The NVIDIA BlueField-4 Advanced Secure Trusted Resource Architecture simplifies network operation, enhances tenant isolation, and delivers greater control and visibility for millions of GPU AI clusters.
Secure AI for AI Factory
The fact is that agentic workflows are dealing with a growing volume of proprietary data, restricted content, and critical mission-critical models to power agentic AI factories. This necessitates advanced infrastructure safety customised for autonomous agents in collaborative or cloud-based environments where infrastructure is not automatically trusted.
The Vera Rubin platform is built on full-stack NVIDIA Confidential Computing for a secure execution environment at rack scale. Vera Rubin NVL72 integrates Rubin GPUs, Vera CPUs, NVIDIA NVLink™ security and networking functions into a single platform for encrypting data through high-speed interconnects.
This offers hardware-level authentication to make sure that the system itself is counterfeit-proof.
NVIDIA Confidential Computing is being embraced by cloud providers Firmus, CoreWeave, GMI Cloud, IBM Cloud, Lambda, IREN, Microsoft Azure, Nebius, Nscale, and SpaceXAI as well as Vultr.
But to provide this level of safeguarding at POD scale needs a programmable software layer that can enforce and orchestrate as well as adapt security protocols throughout the entire system. The NVIDIA DOCA™ software platform provides sophisticated safety for every Vera Rubin platform rack along with an AI factory layer, thereby securing data, agents, and context memory as well as AI inference alongside features implemented directly within BlueField-4 silicon.
DOCA allows for multi-tenant network separation, zero-trust enforcement of policies, runtime threat identification and end-to-end encryption at as much as 800 Gb/s speeds without taxing host CPU resources, therefore enterprises can safely and securely scale AI factories.
Speeding up the Development of AI Factories
The NVIDIA DSX™ platform offers the complete design and operating foundation for Vera Rubin AI factories, bringing together reference design, modelling, standard infrastructure software, facilities, and ecosystem-based technologies so as to help construct and manage energy-efficient AI factories optimized for the lowest token cost.
DSX is designed for the Vera Rubin POD architecture that corresponds to every layer of the stack, right from silicon and systems to lifecycle management as well as multi-tenant operations, significantly speeding up deployment and raising the standard for operational dependability along with resilience at scale.
NVIDIA DSX helps Lenovo, Dell Technologies, HPE, and Supermicro as well as ASUS, Foxconn, Gigabyte, Pegatron, Quanta Cloud Technology – QCT, Wistron, and Wiwynn accelerate AI factory ramp with Vera Rubin.
Notably, the production shipments of Vera Rubin are scheduled to start this autumn.