Skip to main content
Back to Glossary
ai-infrastructure

Vera Rubin

NVIDIA's GPU architecture after Blackwell, planned for 2026, targeting a further ~4× increase in AI compute per chip and continuing the biennial architecture cadence.

Full Definition

Vera Rubin is the codename for NVIDIA's GPU architecture succeeding Blackwell, named after the astronomer who confirmed the existence of dark matter. Announced at GTC 2024, the Vera Rubin platform is expected to ship in 2026, pairing the Vera CPU with the Rubin GPU in a successor to the Grace-Blackwell Superchip.

NVIDIA's published roadmap indicates Vera Rubin will deliver approximately 4× the AI performance of Blackwell per chip, continuing the biennial architecture cadence (Ampere → Hopper → Blackwell → Vera Rubin). Infrastructure planners at AI factory operators must account for Vera Rubin's power density and cooling requirements in their 2025–2027 facility roadmaps, since architectural decisions made today will constrain or enable the next GPU generation.

Also Known As

Rubin GPUNVIDIA Rubin

Source Reference

NVIDIA Jensen Huang GTC 2024 Keynote; NVIDIA Architecture Roadmap (2024)

Related Terms