Nvidia CEO Jensen Huang announced on Monday that the company’s next generation of chips is now in full production. He said the new chips can deliver five times the artificial intelligence computing power of Nvidia’s previous generation when running chatbots and other AI applications, Reuters reported.

Speaking at the Consumer Electronics Show in Las Vegas, Jensen Huang revealed new details about the upcoming chips, which are set to arrive later this year. Nvidia executives told Reuters that the chips are already being tested in the company’s labs by several AI firms.

The announcement comes as Nvidia faces growing competition, not only from traditional rivals but also from some of its largest customers. Despite this, Nvidia continues to dominate the market for training AI models.

At the center of the announcement was the Vera Rubin platform. The platform is made up of six separate Nvidia chips. The flagship server will contain 72 graphics processing units and 36 newly designed central processors. According to Jensen Huang, the chips can be connected into large pods with more than 1,000 Rubin chips working together.

Moreover, Huang said the new system can improve the efficiency of generating AI tokens by up to ten times. Tokens are the basic units used by AI systems to process and generate responses.

However, he noted that achieving this performance depends on a proprietary type of data used by the Rubin chips. Nvidia hopes the broader industry will adopt this approach. Huang said the company achieved a major leap in performance even though the chip design uses only 1.6 times more transistors than the previous generation.

In addition, much of the presentation focused on AI inference. This refers to delivering AI responses to millions of users through chatbots and other applications, an area where competition is increasing from companies such as AMD and Google.

To support this, Nvidia revealed a new storage layer called context memory storage. The technology is designed to help chatbots respond faster to long and complex questions, especially during extended conversations.

Nvidia also announced a new generation of networking switches. These switches use co packaged optics, a technology that helps link thousands of machines into a single system. The offering competes with products from Broadcom and Cisco Systems, Reuters reported.

Furthermore, Nvidia said CoreWeave will be among the first companies to deploy the Vera Rubin systems. The company also expects Microsoft, Oracle, Amazon, and Alphabet to adopt the new platform.

Beyond data centers, Jensen Huang highlighted new software aimed at self driving cars. The software helps autonomous vehicles decide which path to take and creates a record engineers can review later. The software, known as Alpamayo, was shown in research form last year and will now be released more widely.

Huang added that Nvidia will also open source the data used to train the models. He said this approach allows automakers and engineers to better understand and trust how the AI models were developed, according to Reuters.

Related post

View all