Nvidia recently provided further details on its new AI data center computing platform—Vera Rubin—a launch which may have significant repercussions for the future of AI, given the industry’s vast reliance on the firm’s technology.
Previously, Nvidia had revealed some specifics about Vera Rubin, but at the CES tech conference in Las Vegas on Monday, they elaborated on how the system will function and disclosed the rollout timing. Vera Rubin is currently under manufacturing, and initial products based on it will arrive in the latter half of 2026, the company stated.
Nvidia has become emblematic of the AI boom, owing to the widespread adoption of its AI-based chips and platforms, which briefly allowed it to become the world’s first company to achieve a $5 trillion valuation last year. However, the firm is also grappling with concerns about an AI bubble amidst escalating competition and the drive by tech corporations to develop their own silicon to lessen dependency on Nvidia.
Nvidia CEO Jensen Huang, clad in his signature leather jacket, addressed the source of AI funding—key to the bubble debates—in his opening presentation at the Fontainebleau Las Vegas theater. He noted that corporations are shifting their R&D budgets from traditional computational methods to artificial intelligence.
“People ask where the money is coming from? That is where the money is originating,” he remarked.
The Vera Rubin platform represents Nvidia’s bid to position itself as the answer to the computational challenges posed by increasingly demanding AI models—for instance, whether existing infrastructure can handle more intricate AI queries. The firm asserts in a press release that its future AI server rack, dubbed the Vera Rubin NVL72, “delivers more bandwidth than the entire internet.”
Alongside Vera Rubin, Nvidia announced the development of a novel storage system geared to assist AI models in processing more complex, context-aware queries faster and more effectively. Current storage and memory types used by conventional computers and even the GPUs powering data centers will prove insufficient as firms like Google, OpenAI, and Anthropic transition from simple chatbots to full-fledged AI assistants.
Huang spoke about the shift from chatbots to agents on Monday. In a video demonstration, an individual created a personal aide by linking a friendly desktop robot to several AI models running on an Nvidia DGX Spark workstation. The robot could perform actions such as recalculating a user’s to-do list and even instructing a dog to get off a couch.
Huang mentioned that creating such an assistant would have been unthinkable just a few years ago, but now it is “entirely trivial” as developers can leverage large language models instead of legacy tools for building applications and services.
In other words, the old approach simply won’t suffice as AI grows more sophisticated and “undertakes” tasks requiring multiple steps, Nvidia contends.
“The bottleneck is shifting from the compute system to context management,” Dion Harris, Nvidia’s Senior Director of HPC and Hyperscale AI Solutions, commented to journalists before the press conference.
“Storage can no longer be secondary,” he added.
Nvidia also recently secured a licensing accord with inference specialist Groq shortly before CES—another indicator of substantial investments in this AI segment.
“Inference is now a thought process, rather than a one-shot response,” Huang stated, referring to the procedure AI models go through to “deliberate” and “reason” through their outputs and execute tasks.
All major cloud providers—Microsoft, Amazon Web Services, Google Cloud, and CoreWeave—will be among the first to adopt Vera Rubin, according to Nvidia’s release. Computing companies like Dell and Cisco are anticipated to integrate the new chips into their data centers, while AI labs such as OpenAI, Anthropic, Meta, and xAI are likely to embrace the new technology for training and delivering more sophisticated responses.
Nvidia also bolstered its autonomous vehicle efforts with new Alpamayo models and “physical AI”—a form of AI that controls robotics and other real-world machinery, building upon the vision laid out at the GTC conference in October.
However, Nvidia’s progress and prevalence also mean it shoulders the burden of continuously exceeding Wall Street’s high forecasts, easing worries that AI infrastructure spending is vastly outpacing actual demand.
Meta, Microsoft, and Amazon, among others, have shelled out tens of billions in capital expenditure this year alone, and McKinsey & Company projects that companies will invest nearly $7 trillion in worldwide data center infrastructure by 2030. And much of the support poured into AI appears tied to a relatively small cohort of corporations trading funds and technology in what is termed “circular funding.”
About the Author
