Key facts
- Nvidia CEO Jensen Huang confirmed Samsung, SK Hynix, and Micron are certified to supply HBM4 memory.
- The HBM4 memory will be used for Nvidia's Vera Rubin AI accelerator platform.
- Vera Rubin is expected to deliver 10x agent throughput compared to the Grace Blackwell platform.
- Deliveries for Vera Rubin are scheduled to begin in Q3 2026.
- SK Hynix is estimated to supply 60-70% of HBM4 volume for Vera Rubin.
- Nvidia is opening a new research and development center in South Korea.
Nvidia CEO Jensen Huang announced on June 5, 2026, that Samsung, SK Hynix, and Micron have all passed certification to supply HBM4 high-bandwidth memory for the company's Vera Rubin AI accelerator platform. Vera Rubin, the successor to the Grace Blackwell architecture, has entered full production and is expected to deliver 10 times the agent throughput. Deliveries are slated to begin in Q3 2026. Supply chain analysts estimate SK Hynix will supply 60-70% of the HBM4 volume, Samsung 25-30%, and Micron the remainder. Huang made the announcement during a visit to South Korea, where he also met with leaders from SK Group, Samsung, LG Group, Hyundai Motor Group, and Naver to discuss supply commitments and AI partnerships. Nvidia is also establishing a new R&D center in South Korea. Despite the positive news for Nvidia and its memory partners, Micron's stock fell 7.74% on June 5, influenced by broader tech sector weakness and Broadcom's earnings report. Separately, SK Hynix was reportedly already planning to double its memory production capacity by 2030-2031, with its Chinese facilities playing a key role in this expansion, even before Huang's informal request for more memory at Computex.