Nvidia's Vera Rubin Chips Enter Full Production, Says Jensen Huang

During a major press event at the CES technology show in Las Vegas on Monday, Jensen Huang shared exciting news regarding Nvidia's latest AI platform. The new Vera Rubin superchip platform is on track to begin shipping to customers within the year. 'As of today, Vera Rubin is fully in production,' Huang announced.

According to Nvidia, the Rubin platform is set to slash the costs associated with AI training and operations to merely a tenth of what their top-tier systems currently require. Additionally, Rubin can train certain sizable models using only about a quarter of the chips needed by Blackwell. These advancements could lower expenses significantly for advanced AI systems, making Nvidia's hardware hard to surpass.

Nvidia has revealed that pioneering partners, Microsoft and CoreWeave, will soon launch services utilizing Rubin chips, expected later this year. Nvidia reports that new data centers being built by Microsoft in Georgia and Wisconsin will accommodate thousands of Rubin units. Additionally, some partners are already deploying their next-gen AI models using preliminary Rubin systems.

The semiconductor giant is also working with Red Hat to explore further ventures leveraging the power of the Rubin chips. Red Hat, known for its open-source software solutions for sectors like banking, automotive, airline, and government, is a strategic partner in this initiative.

Named after the astronomer Vera Rubin, this latest chip platform features a comprehensive array of six distinct chips, including the Rubin GPU and Vera CPU. These components are crafted utilizing TSMC's 3-nanometer technology and the most current high-bandwidth memory techniques, seamlessly interconnected with Nvidia's sixth-generation interconnect and switching technologies.

At the CES conference, Huang described each element of the Rubin chip platform as 'truly revolutionary and unrivaled in its category.'

The journey of developing the Rubin system has been a long one for Nvidia, with earlier announcements taking place in 2024. Last year, Nvidia indicated that Rubin-based systems would become available starting in the latter half of 2026.

While 'full production' is currently proclaimed, specifics regarding this status remain somewhat unclear. Typically, sophisticated chip manufacturing starts on a smaller scale for testing and validation before scaling up considerably later.

Austin Lyons, a Creative Strategists analyst, views Nvidia's CES statement as an assurance to investors of timely progress. Despite earlier speculation on Wall Street about potential production delays with the Rubin GPU, today's declaration affirms Nvidia's completion of essential development phases, promising a production ramp-up by the latter part of 2026.

It's notable that Nvidia previously faced setbacks with its Blackwell chips in 2024 due to a design problem that led to overheating in server setups, which postponed shipments to mid-2025.

Given the burgeoning AI industry, there's a fierce race among software firms and cloud providers to secure access to Nvidia's latest GPUs. The demand for Rubin is expected to be similarly robust, though some companies are diversifying their strategies by creating their own chip architectures. OpenAI, for instance, is collaborating with Broadcom to develop customized silicon for its next-gen AI initiatives, reflecting a potential long-term challenge for Nvidia.

Nonetheless, Lyons suggests Nvidia's expansion beyond GPUs to becoming a comprehensive AI system architect is significant. Offering integration across computing, networking, memory, storage, and software orchestration, the platform presents itself as a formidable competitor, even as some major players opt for custom solutions.