The chip giant is preparing new "Rubin" and "Feynman" architectures to help companies address generative AI's skyrocketing costs and energy demands.

Rubin will be followed by Rubin Ultra in the second half of 2027, a collection Vera Rubin GPUs, Huang said. Around the same time, Nvidia will release a Rubin Next chip that combines four GPU dies to make a single chip, doubling Rubin’s speed.

This year’s Blackwell Ultra isn’t what we originally expected, when Nvidia said last year that it would begin producing new AI chips on a yearly cadence, faster than ever before. But Nvidia quickly moved on from Blackwell Ultra during today’s GDC keynote to reveal its next architecture,