Dive Brief:
- Nvidia released pivotal updates for enterprises focused on data centers and compute at the CES 2026 event in Las Vegas. On Monday, the company launched its latest computing architecture, the Rubin platform. Neocloud provider CoreWeave, which counts IBM and OpenAI as two of its customers, will be one of the first vendors to offer the platform.
- The Rubin platform uses six chips to support building and deploying advanced AI systems that the company said delivers cheaper inference results and uses fewer GPUs for model training than its predecessor, the Nvidia Blackwell platform. The anticipated reduction in training resources and inference costs will help “accelerate mainstream adoption,” the chip giant said in a press release.
- “Vera Rubin is designed to address this fundamental challenge we have: The amount of computation necessary for AI is skyrocketing; the demand for Nvidia GPUs is skyrocketing,” Nvidia CEO and Founder Jensen Huang said in a keynote at CES. “It’s skyrocketing because models are increasing by a factor of 10 in an order of magnitude every single year.”
Dive Insight:
Rising compute demand is something tech giants became intimately familiar with in 2025 as enterprises rushed to adopt and deploy new AI tools.
Microsoft reported in a Q1 2026 earnings call last fall that it was facing a compute capacity shortage that will affect the company throughout its fiscal year. Indeed, increasing AI workloads have prompted nearly 80% of organizations to consider their AI data center needs a year in advance, according to a report from IT services management company Flexential.
Microsoft, along with AWS, Google, Oracle and OpenAI, are expected to utilize Nvidia’s Rubin platform amid the capacity crunch. It’s not just hyperscalers or large AI model developers interested in the technology; Nvidia’s press release included traditional IT players Dell, HPE and Lenovo touting the news as well.
The Rubin platform is designed to address the reality that “next generation AI factories” must process thousands of input tokens to provide context for agentic reasoning and complex workflows while simultaneously maintaining real-time inference under power, cost and deployment constraints, said Kyle Aubrey, director of technical marketing for the accelerated computing product team at Nvidia, in a blog post. AI factories are specialized infrastructure stacks designed to manage and streamline the AI lifecycle.
Different components — including GPUs, CPUs, power delivery and cooling structures — were built together as a single system to make up the foundation of the Rubin platform, Aubrey said.
“By doing so, the Rubin platform treats the data center, not a single GPU server, as the unit of compute,” Aubrey said. “This approach establishes a new foundation for producing intelligence efficiently, securely and predictably at scale.”
Nvidia wasn’t the only player to unveil a rack-scale platform at CES. AMD also introduced Helios, which is designed to “deliver maximum bandwidth and energy efficiency for trillion-parameter training,” according to a press release.
Compute infrastructure serves as the foundation for AI, which is driving unprecedented global compute capacity expansion, the AMD release said.
“AMD is building the compute foundation for this next phase of AI through end-to-end technology leadership, open platforms, and deep co-innovation with partners across the ecosystem,” AMD CEO and Chair Lisa Su said in the release.
