Nvidia will launch a new AI chip late this year. The company says it will consume less energy than earlier-generation chips while delivering strong computing performance.

Nvidia CEO Jensen Huang on Jan. 5 local time introduced the Vera Rubin next-generation chip, developed over three years, at CES 2026 in Las Vegas.

Huang said Vera Rubin is focused on handling AI requests faster and more cheaply than earlier-generation chips. He said AI model training is possible using only a quarter of Blackwell, Nvidia's current flagship GPU, and that it can provide information needed for chatbots and other AI products at about one-tenth of existing costs.

A design that reduces the number of cables is also cited as a strength, allowing faster installation in data centres.

Vera Rubin will be shipped in the second half to customers including Microsoft and Amazon, the New York Times reported. Huang first mentioned Vera Rubin in March last year through a conference.

He also referred at this CES to Nvidia's autonomous driving-related project under ambitious development. Mercedes-Benz plans to start shipping vehicles this year equipped with Nvidia autonomous driving technology that responds to Tesla Autopilot.