The new Vera Rubin product is a major shift from the previous type of AI architecture that launched in late 2024, according to Nvidia.
The model, named after an American astronomer, is said to be five times more efficient and will be available in the second half of 2026. It consists of six chips that, the company says, form an AI supercomputer.
Today, Nvidia has an estimated 80 percent of the global AI data center chip market, but it faces challenges in maintaining its leading position.
Several other chip manufacturers are working hard to compete in the market, and some of Nvidia's largest customers - Google, Amazon and Microsoft - are increasingly developing their own chips to reduce their dependence on the company.
China is also trying to build its own alternatives to Nvidia's products.
