Part 2/9:
In tandem with the announcement of the Vera Rubin GPU, Huang introduced its future counterpart, the Vera Rubin Ultra, anticipated for release in 2027. The new series advances NVIDIA’s commitment to spearheading AI technology. With 50 petaflops of FP4 inference performance per chip when installed in fully equipped NVL 144 rack systems, the Vera Rubin GPU is projected to deliver a remarkable 3.6 exaflops of FP4 inference compute—over three times the power of the previous Blackwell Ultra configuration.