Nvidia CEO Jensen Huang confirmed that the upcoming Vera data center central processing units will utilize DRAM and HBM4 memory chips from SK Hynix. This announcement is part of a broader, multi-year partnership between the two companies to co-develop next-generation memory for Nvidia’s AI supercomputer platforms.
Key Highlights of the Deal
- Vera & Rubin Platforms: SK Hynix is codeveloping and supplying advanced memory for the Vera Rubin supercomputing architecture and standalone Vera CPUs.
- Advanced Memory Tech: The chips will utilize the next-generation HBM4 (High Bandwidth Memory) and second-generation LPDDR5X subsystems to achieve speeds up to 1.2 TB/s while keeping energy draw under 30W.
- Diversified Supply Chain: While SK Hynix remains its largest partner, Nvidia has also qualified and initiated production with Samsung Electronics and Micron Technology to meet massive demand.
- AI Chip Manufacturing: The multi-year deal extends to chip design. SK Hynix will utilize Nvidia’s software, including CUDA-X, Omniverse, and cuOpt, to build digital twin factories and optimize semiconductor fabrication.
The Vera CPU represents Nvidia’s aggressive push into the standalone data center CPU market, directly challenging incumbents like Intel and AMD by prioritizing high-speed agentic AI tasks.
Would you like to know more about how Nvidia’s Vera CPUs stack up against traditional x86 processors, or are you interested in how HBM memory scaling impacts AI computing performance? Let me know how you’d like to proceed!
