NVIDIA Rubin unveiled at CES 2026: Is NVIDIA Rubin the AI chip that will redefine global data centres?

At CES 2026, NVIDIA unveiled Rubin, its next-generation AI supercomputer platform designed for large-scale data centres, combining GPUs and CPUs into a single extreme-codesigned system. NVIDIA announced its Rubin platform, named after astronomer Vera Rubin, and branded as the first "extreme-codesigned" AI system with six new chips operating together as a single giant brain. In a two-hour keynote, CEO Jensen Huang said that the speed of computing and AI has entirely transformed the way computers are operated with Rubin being the next breakthrough to developing smarter machines.

What NVIDIA Rubin brings to AI computing

The NVIDIA Rubin AI chip platform integrates multiple GPUs and CPUs into a rack-scale system built for high-performance AI workloads. Designed specifically for NVIDIA Rubin data centres, the Vera Rubin NVL72 rack combines 72 Rubin GPUs and 36 Vera CPUs into a single AI data center chip architecture. Rubin integrates 72 high-performance Rubin GPUs and 36 Vera CPUs in a rack known as Vera Rubin NVL72, a supercomputer system to be implemented in data centers. Every Rubin GPU consists of 336 billion transistors, or tiny switches that render the chips smart and fast, whereas every Vera CPU consists of 88 custom cores with 227 billion transistors. NVIDIA claims that this system reduces the AI inference costs by up to 10 times than the previous Blackwell chips, which means that companies spend less money to run AI models such as chatbots or image generators.

How NVIDIA Rubin the chip tech is better

Positioned as the NVIDIA Blackwell successor, Rubin reduces AI inference costs by up to 10x, making large-scale AI deployment more affordable for enterprises and cloud providers. Rubin enhances its previous chips by closely working together in a joint venture between its GPUs and CPUs that were initially intended to work together in the field of AI.

The GPUs perform large mathematical computations to train very large AI models, and Vera CPUs do complex instructions using their Olympus cores, and exchange data with each other at an unprecedented speed, without bottlenecks. The more transistors the faster Rubin can process larger data sets, with less power consumption, and scalable easily - imagine Lego blocks stacked to create planet-sized artificial intelligence without frying servers through heat.

By combining Rubin GPUs with Vera CPUs, NVIDIA is building an AI supercomputer for data centres that enables faster model training, real-time inference, and lower power consumption. The platform will make AI more affordable and quicker to all, whether you are a company creating a chat app or even a giant like Google that is operating a search engine or a self-driving car.

Reduced inference fees allow small companies to operate advanced AI without enormous invoices and accelerate technologies, such as real-time translation or medical scans. Rubin continues to position NVIDIA higher in the AI race than its competitors such as AMD, so that by the end of 2026 all data centers across the globe will be upgraded to support smarter and more efficient AI.

Open models for all 

NVIDIA’s open models — trained on NVIDIA’s own supercomputers — are powering breakthroughs across healthcare, climate science, robotics, embodied intelligence and autonomous driving.

“Now on top of this platform, NVIDIA is a frontier AI model builder, and we build it in a very special way. We build it completely in the open so that we can enable every company, every industry, every country, to be part of this AI revolution.”

The portfolio spans six domains — Clara for healthcare, Earth-2 for climate science, Nemotron for reasoning and multimodal AI, Cosmos for robotics and simulation, GR00T for embodied intelligence and Alpamayo for autonomous driving — creating a foundation for innovation across industries.

“These models are open to the world,” Huang said, underscoring NVIDIA’s role as a frontier AI builder with world-class models topping leaderboards. “You can create the model, evaluate it, guardrail it and deploy it.”

The road ahead

Following its CES 2026 AI chips announcement, NVIDIA Rubin systems are expected to ship to data centres through partners like Dell and HPE in the second half of 2026. Huang explained that NVIDIA builds entire systems now because it takes a full, optimised stack to deliver AI breakthroughs. “Our job is to create the entire stack so that all of you can create incredible applications for the rest of the world,” he said.

Rubin will now start to produce, and partner products such as Dell or HPE servers will enter markets in the second half of 2026. Anticipate cloud providers such as AWS and Azure to be offering Rubin-powered experiences, reducing prices on AI apps to the point of creating videos or code 10x cheaper. By 2027, developers will train models 100x larger than current models (which will enable new breakthroughs in drug discovery or climate modeling), and everyday users will experience even faster AI assistants on phones and PCs.