Nvidia’s Rubin on the Horizon: Successor to Blackwell Arrives in 2026
At the Computex trade show in Taipei, Taiwan, Jensen Huang, CEO of Nvidia, revealed the name of their next-generation chip architecture during his keynote on Sunday evening. “Rubin,” named after famed astronomer Vera Rubin, will succeed the current “Blackwell” line of GPUs and CPUs. Scheduled for launch in the first half of 2026, this announcement underscores Nvidia’s commitment to releasing new chip architectures annually. With a hint of playful apprehension, Huang joked to the audience right before unveiling the Rubin line, “This is the first time anyone will see this, and I’m not sure if it’s a good idea or not.”
“All of these chips are in full development, 100% of them, and the rhythm is one year, at the limits of technology, all 100% architecturally compatible,” said Huang, “and all of the richness of software on top of it.”
Nvidia’s Rubin reveal follows closely behind their official announcement of the Blackwell architecture at the GTC conference just three months ago. Research firm Jefferies & Co. suggests the Rubin chips will incorporate the next-generation memory standard, HBM4, replacing the current top-of-the-line HBM3e. This applies to both the GPU family (designated R100 GPUs) and a central processing unit (CPU) codenamed Vera.
Jefferies analysts predict Rubin chips will be manufactured by TSMC using their cutting-edge 3-nanometer process, with production slated for late 2025. The design utilizes a “4x reticle design” compared to Blackwell’s 3.3x, meaning it occupies more space on a silicon wafer during the chip creation process. Additionally, Huang hinted at a future “Blackwell Ultra” GPU alongside a confirmation of a “Rubin Ultra” architecture arriving in 2027.
