NVIDIA's 2026 AI breakthrough

At the Computex conference in Taipei, NVIDIA CEO, Jensen Huang, announced ‘Vera Rubin’, its next-generation AI chip and Blackwell chip successor (which it just released a few months ago), due to be launched in 2026, following the release of Blackwell Ultra, next year.

What is the Vera Rubin AI chip?

The Vera Rubin AI chip was named after Vera Florence Cooper Rubin, a famous astronomer who established the presence of dark matter.

It will have new graphics processors (GPUs), central processors (CPUs), and networking chips designed to power AI applications with high-bandwidth memory which will be supplied by manufacturers such as Samsung.

Why have they announced this?  

NVIDIA plans to upgrade its AI chips every year, as opposed to every two years (as they did previously) because Huang believes that “we are seeing computation inflation.”

He said that traditional computing methods can't process the amount of data needed to power AI machines, so users need accelerated computing power–like NVIDIA’s chip platforms–to cut costs and increase efficiency, stating that NVIDIA chips will create 98% cost savings and use 97% less energy.  

The announcement also highlights NVIDIA’s aggressive plans to hold onto its 80% share of the AI chip market, for as long as they can as more and more competitors move into the space, as demand for high-performance AI chips increases.