Nvidia on Monday said its Vera Rubin agentic AI platform and Vera CPUs are in full production and on schedule to ship this fall, and it introduced DSX OS, a new operating system designed to help organizations manage and operate AI factories more efficiently. The announcements were made during CEO Jensen Huang’s keynote at the GTC Taipei conference.. Nvidia unveiled DSX OS, a modular, open source software for AI factory operators to provision, operate, and monitor AI infrastructure at scale. The company said DSX OS provides lifecycle management, health automation, resiliency, and multi-tenant operations.. “It’s a stable foundation for which new services, agents, and AIs can be built,” said Ian Buck, Nvidia’s vice president of hyperscale and high-performance computing, during a media briefing.. Nvidia also introduced DSX MaxLPS, a suite of technologies designed for Nvidia’s next-generation Vera Rubin AI platform to maximize token performance within a fixed power budget. “With MaxLPS, AI factories can safely deploy up to 40% more GPUs within the same power envelope,” Buck said. “That’s 40% more compute, 40% more tokens, and 40% more revenue than was possible before.”. Related:NVIDIA Unveils GPUs to Power Clouds. Nvidia’s announcements expand its hardware and software portfolio to help cloud providers, AI companies, and enterprises build and operate AI infrastructure, and to provide tools for building and running AI agents and other AI applications.. In his keynote, Huang said Nvidia is more than a chip-and-systems company.. “Nvidia has really become an infrastructure company, not just a GPU company, not just a systems company, but an infrastructure company to help you generate the maximum revenues, the maximum profit, and to get there as soon as possible,” Huang said.. Nvidia first announced its DSX platform, Vera Rubin AI platform and Vera CPU at its GTC conference in San Jose in March. The company has previously introduced a DSX reference design to guide operators in building AI factories and a DSX blueprint for creating physically accurate digital twins of those facilities.. According to Nvidia, the Vera Rubin platform, described as the company’s most comprehensive AI system to date, combines five rack-scale systems into a single AI supercomputer: NVL72 GPU racks (featuring 72 Rubin GPUs and 36 Vera CPUs), Vera CPU racks, Groq 3 LPX inference accelerator racks, BlueField-4 STX storage racks, and Spectrum-6 SPX Ethernet networking racks.. Huang Pitches Vera Rubin and Vera CPU. In his keynote, Huang once again touted Nvidia’s forthcoming hardware as built for agentic AI, calling Vera Rubin “the most ambitious endeavor in the history of our company” and saying Vera CPU will be Nvidia’s next major growth driver. “All of the CPUs until now were created for people,” Huang told the audience. “This CPU was built for agents.”. Related:Nvidia Brings Blackwell GPUs to Enterprise Data Centers. Huang said agents need much faster response times, which Vera delivers. In a new benchmark, the company said that Vera CPUs deliver 1.8 times the performance of x86 CPUs. “Agents are impatient. They don’t live in a world that is in seconds. They live in a world that’s in nanoseconds,” he said. “It is vital that we make the CPUs as low-latency as possible. So we created Vera CPU for the age of AI.”. Huang added that Vera orders will make it “the fastest and the most successful product launch in our company’s history.” In fact, Nvidia on Monday announced that Anthropic, OpenAI, and SpaceXAI are among the early adopters of the Vera CPU. Other customers include Oracle Cloud Infrastructure, ByteDance, and neoclouds CoreWeave, Nebius, Nscale, and Lambda.. Nvidia’s ‘Customer Traction’. Matt Kimball, vice president and principal analyst at Moor Insights & Strategy, told Data Center Knowledge that the most significant part of the Vera announcement is the customer traction. He said that Oracle Cloud’s support for the Vera Arm-based CPU is notable given Oracle’s longstanding use of the Ampere Arm CPU.. Related:Nvidia CEO Shares Vision for Overhauling Data Centers. “Nvidia has real customers, big customers, real deployments, and new logos for Vera, if you will,” Kimball said.. Hardware makers building standalone Vera CPU systems include Dell Technologies, HPE, Lenovo, and Supermicro. The Vera-powered systems, with expected availability in the third quarter, will be offered as liquid-cooled racks for large-scale agentic AI and reinforcement learning workloads and as two-socket, air-cooled systems for enterprise, cloud, data processing, and AI factory deployments, the company said.. Nvidia said Vera is designed to “drive diverse workloads across industries,” including agentic AI, reinforcement learning and data processing. To Kimball, the language is deliberate. He believes the company is trying to establish Vera as not only a CPU for AI workloads but for a broad range of enterprise tasks. He believes Nvidia is positioning itself to compete against AMD Epyc and Intel Xeon CPUs in the long term.. “They’re trying to position Vera as being very good at AI, but also able to support a broader swath of workloads than just AI,” he said. “That tells me they definitely have a sight toward the enterprise.”. As for Vera Rubin, system builders, and software and storage partners include Dell, HPE, Lenovo, Supermicro, Hitachi Vantara, IBM, Nutanix, NetApp, and VAST Data, the company said. They will begin shipping this fall.. As for DSX OS and DSX MaxLPS, Kimball said the announcements address one of the biggest challenges organizations face when deploying AI: how to deploy the right infrastructure for their needs and use those resources efficiently as they scale. DSX OS is designed to better manage the lifecycle, job scheduling, and resource utilization of that infrastructure.. “When you take that from a few resources to thousands to tens of thousands of resources, you start to lose a lot of efficiency at scale, and that is what DSX OS is trying to solve for,” he said.. Nvidia’s Other Announcements. Nvidia also made several other announcements on Monday:. Nvidia DGX Station is a deskside AI supercomputer for developing and running agents on Windows.. New Nvidia Agent Toolkit software, including Nvidia NemoClaw blueprints, Nemotron models, OpenShell secure runtime, and CUDA-X libraries with agent skills.. New open source agent tools and skills for physical AI.. TSMC is using Nvidia’s accelerated computing and AI to advance semiconductor design and manufacturing.. Foxconn is deploying Nvidia AI in Taiwan’s leading medical centers.
