Nvidia Chief Executive Jensen Huang said on Monday that the company’s next generation of chips is now in full production, promising a fivefold increase in artificial intelligence computing power compared with its previous offerings. Speaking at the Consumer Electronics Show in Las Vegas, Huang unveiled new details about the chips, which are set to arrive later this year and are already being tested by AI firms in Nvidia’s laboratories.
The new platform, named Vera Rubin, consists of six separate Nvidia chips and is designed to power chatbots and other AI applications at massive scale. The flagship configuration includes 72 of Nvidia’s graphics processing units alongside 36 newly designed central processors. Huang demonstrated how the systems can be linked together into large “pods” containing more than 1,000 Rubin chips, underscoring Nvidia’s focus on serving the biggest AI workloads.
Huang said the performance gains were achieved partly through the use of a proprietary data format that Nvidia hopes will be adopted more broadly across the industry. Despite having only about 1.6 times as many transistors as its predecessor, the Rubin platform delivers a dramatic leap in performance, he said, calling it a “gigantic step up” for AI computing.
While Nvidia remains the dominant force in training large AI models, competition is intensifying in the deployment of those models to consumers. Traditional rivals such as Advanced Micro Devices are vying for market share, while major customers including Alphabet’s Google are increasingly developing their own chips to reduce reliance on Nvidia. Much of Huang’s keynote focused on how the Rubin chips are optimized to deliver fast, efficient responses for chatbots used by hundreds of millions of people simultaneously.
To support that goal, Nvidia introduced a new layer of storage technology called “context memory storage,” designed to help AI systems handle long conversations more smoothly under heavy user demand. The company also announced a new generation of networking switches using co-packaged optics, a technology critical for connecting thousands of machines into a single computing system and one that pits Nvidia against competitors such as Broadcom and Cisco Systems.
Beyond data centers, Nvidia highlighted advances in software for self-driving cars. Huang said the company would more broadly release its Alpamayo software, which helps autonomous vehicles make driving decisions while leaving a detailed record engineers can review later. Nvidia also plans to open-source both the models and the data used to train them, a move Huang said is essential for building trust in AI systems.
The announcements come as Nvidia sharpens its competitive edge through acquisitions and talent hires. Last month, the company absorbed chip technology and executives from startup Groq, including figures who played key roles in Google’s in-house AI chip development. Although Google remains a major Nvidia customer, its custom chips are increasingly seen as a significant threat to Nvidia’s AI stronghold.
Nvidia is also eager to distinguish its latest products from older chips such as the H200, which U.S. President Donald Trump has allowed to be sold to China. That chip, a predecessor to Nvidia’s current Blackwell platform, has seen strong demand in China, raising concerns among U.S. lawmakers. Against that backdrop, Huang’s message in Las Vegas was clear: Nvidia’s next generation is designed to stay well ahead of both rivals and restrictions.
