Nvidia Chief Executive Officer Jensen Huang on the 5th (local time) unveiled the actual next-generation artificial intelligence (AI) chip "Vera Rubin," saying it "has gone into mass production." He also said the Autonomous Driving car developed with Mercedes-Benz is slated for release in the first quarter of this year.
Huang took the stage as a speaker at the "Nvidia Live" event held at the Fontainebleau hotel in Las Vegas, saying, "We are mass producing Vera Rubin NVL72." The product boasts higher performance than the currently available AI chip "Grace Blackwell" (GB). It combines 36 "Vera" central processing units (CPUs) and 72 "Rubin" graphics processing units (GPUs) into one. Compared with existing products, inference performance is five times higher, and the expense per token has been reduced to one-tenth.
Held a day before the opening of the world's largest IT and home appliance exhibition, "CES 2026," the event took place at the 3,600-seat BleauLive Theater, and all seats were filled about 15 minutes before Huang's speech began. Nvidia therefore broadcast Huang's CES 2026 special address live on a large electronic display set up at another facility in the hotel. That venue also quickly filled with spectators.
Huang said, "We believed we had to advance computing technology every year without falling behind by even one year," adding, "To do that, we need to be entering production around now. Vera Rubin is currently in full mass production." He added, "As computational demand is surging in both AI training and inference, we are bringing Vera Rubin to market at the right time."
Nvidia first unveiled the Rubin platform at its annual developer conference, GTC, in Mar. last year. Huang said products applying the Rubin platform will go on sale as planned starting in the second half of this year.
Huang also went on to explain each component that makes up the Vera Rubin product. Regarding the Vera CPU, composed of 227 billion transistors, he said, "Its performance per watt is about twice as good as the most advanced CPUs in the world." The Rubin GPU, suited for AI inference tasks, is composed of 336 billion transistors.
Huang said, "Rather than using the biggest and smartest AI model for every problem, it's better to run the right one for the situation," adding, "The next-generation Rubin platform will function as an AI supercomputer that supports all computation." He continued, "Nvidia is reinventing the entire AI stack, from chips to infrastructure, models, and apps," and said, "We will provide the foundation for our world's developers to build innovative apps."
Huang also unveiled the Autonomous Driving vehicle platform "Alphamayo." It is a platform linked to the foundation model "Cosmos," unveiled at CES 2025 early last year. Huang said its hallmark is the ability not only to recognize facts captured by cameras but also to infer and act on what will happen in the future. Alphamayo will be installed first in Mercedes-Benz's vehicle "CLA." The car will begin its U.S. launch in the first quarter of this year, with sales expanding to Europe and Asia.
Nvidia also unveiled the Robotics model "GROOT." It showed robots from Boston Dynamics and Figure AI operating through it. Huang said, "Nvidia is not just a company that sells chips; it is a place that builds entire systems," adding, "We will help the world participate in the AI revolution."