Nvidia’s ever-optimistic CEO, Jensen Huang, recently unveiled details about the company’s groundbreaking next-gen AI platform at the GTC event. Referred to as the Vera Rubin “superchip,” this processor boasts an astonishing six trillion transistors, representing a vast improvement over existing technologies.

In comparison to the Nvidia RTX 5090 gaming GPU, the Vera Rubin superchip possesses an astounding 60 times more transistors. Additionally, it delivers 100 petaflops of raw computational power—100 times greater than that of Nvidia’s current DGX Spark AI box. Moreover, it features 88 custom Arm cores supporting 176 software threads, facilitating conventional CPU tasks as well.

The characteristics of this chip raise questions about its construction: multiple chip packages contribute to the overall transistor count. The Vera Rubin architecture includes one CPU and two GPU packages, with each GPU package containing two GPUs, culminating in a total of four GPUs.

There’s also potential for the custom-designed CPU cores to make their way into future PC processors. The announcement signifies Nvidia’s ambitious plans, with production set to commence potentially as early as next year, promising an impressive leap in AI technology and performance.