At the GTC 2025 conference, NVIDIA's CEO, Jensen Huang, took the stage to officially introduce the next-generation Vera Rubin superchip to the world for the very first time. The motherboard of this cutting-edge chip seamlessly integrates one Vera CPU and two sizable Rubin GPUs, boasting an impressive array of up to 32 LPDDR memory slots.

In the GPU department, it leverages the power of HBM4 high-bandwidth memory. Huang divulged that the Rubin GPU has already progressed to the laboratory testing phase, with TSMC (Taiwan Semiconductor Manufacturing Company) taking charge of manufacturing the initial batch of samples. Each GPU comes equipped with eight HBM4 interfaces and houses two GPU core chips, each matching the size of a photomask.

The Vera CPU is no slouch either, packing 88 custom-designed Arm architecture cores and capable of supporting up to 176 threads simultaneously. Based on the company's roadmap, the Rubin GPU is projected to enter mass production in either the third or fourth quarter of 2026, marking a significant milestone in the evolution of high-performance computing.