Jensen Huang Unveils Rubin Platform with HBM4 Memory to Succeed Blackwell
NVIDIA has officially opened the next chapter in AI computing. On January 21, CEO Jensen Huang launched the Rubin platform, a powerful successor to the Blackwell architecture. Designed to power the next generation of "AI Factories," Rubin promises to redefine data center performance.
The spotlight is on the technical leap: the new Rubin GPUs are paired with the proprietary Vera CPU and utilize cutting-edge HBM4 memory. This combination delivers a staggering 5x increase in AI inference performance compared to previous models. With memory bandwidth hitting 22 TB/s, NVIDIA aims to solve the critical data bottlenecks faced by developers training massive Large Language Models (LLMs).
This launch is also a strategic move to fend off competition from AMD’s Instinct series. By pushing the boundaries of power efficiency and density, NVIDIA ensures its hardware remain the default choice for global hyperscalers and tech giants in 2026.
Conclusion from HYPERPC:
While Rubin is currently designed for enterprise-level supercomputing, its technology paves the way for the future of consumer graphics. The advancements in memory architecture and processing efficiency will eventually power the next generation of workstations and gaming PCs. At HYPERPC, we are committed to integrating these top-tier technologies into our custom builds as soon as they translate to the professional and enthusiast markets, ensuring our users always stay ahead of the curve.
Verified Sources:
- NVIDIA Blog: Inside the Rubin Platform
- Tom’s Hardware: NVIDIA Rubin Specs and Launch Details
- HPCwire: Rubin vs Blackwell Performance
Egor Streletskiy
5