Key Takeaways
- Jensen Huang, NVIDIA’s CEO, announced that Samsung, SK Hynix, and Micron have successfully qualified to deliver HBM4 memory for the Vera Rubin AI accelerator platform.
- The Vera Rubin platform has transitioned to full production status, with shipments slated for the third quarter of 2026, offering 10x enhanced agent throughput versus Grace Blackwell.
- Industry analysts project SK Hynix will command 60–70% of Vera Rubin’s HBM4 supply, Samsung approximately 25–30%, and Micron the remaining portion.
- Micron shares fell 7.7% on June 5 despite certification confirmation, impacted by tech sector volatility triggered by employment data and Broadcom’s quarterly results.
- The announcement occurred during Huang’s South Korean tour, featuring strategic discussions with executives from SK, Samsung, LG, Hyundai, and Naver regarding AI collaboration and supply capacity.
On June 5, 2026, NVIDIA’s chief executive Jensen Huang publicly validated that Samsung, SK Hynix, and Micron—the trio of dominant memory chip manufacturers—have successfully completed qualification processes to provide HBM4 high-bandwidth memory for the company’s forthcoming Vera Rubin AI accelerator system.
NVDA shares reached $218.66, climbing 1.82% during trading, while Micron (MU) experienced a 7.74% decline.
Upon landing in Seoul, Huang addressed journalists with the confirmation: “All three vendors have been qualified. All three vendors are in production, and they’re all racing to support Vera Rubin.”
The Vera Rubin architecture represents NVIDIA’s next evolution beyond the Grace Blackwell GPU framework. Production ramped to full capacity following Huang’s GTC Taipei keynote presentation on June 1, 2026, with NVIDIA projecting 10x improvement in agent throughput relative to Grace Blackwell capabilities.
“Agentic AI is a new kind of workload,” Huang explained. “One prompt can launch a thousand-step journey of reasoning, retrieval, tool use and response generation. Vera Rubin was built for this moment.”
The platform is scheduled to begin shipping during the third quarter of 2026. Vera Rubin server configurations integrate NVIDIA’s Vera central processors with Rubin graphics units and multiple terabytes of HBM4 memory architecture.
This represents Huang’s initial public validation that all three memory manufacturers have achieved certification standards for the platform, resolving extended uncertainty within supply chain circles.
Projected HBM4 Supply Distribution Among Manufacturers
Although NVIDIA hasn’t published official allocation figures, semiconductor supply chain experts forecast SK Hynix will secure 60–70% of HBM4 volume for Vera Rubin systems. Samsung is projected to capture approximately 25–30% of supply, leaving Micron to fulfill the remaining demand.
SK Hynix initiated qualification procedures before its competitors. Samsung commenced HBM4 volume manufacturing in February 2026. On June 2, Huang publicly called on SK Hynix to accelerate production scaling, emphasizing continued constraints in worldwide semiconductor availability.
NVIDIA is simultaneously establishing a research and development facility in South Korea, with active recruitment underway for the new center.
Micron Shares Decline Despite Qualification Approval
The supplier certification announcement failed to provide support for Micron shares. MU dropped 7.74% on June 5, caught in broader technology sector turbulence following robust U.S. employment figures that unsettled rate-sensitive technology holdings.
The downturn also followed Broadcom’s Wednesday evening earnings release, which appeared to weigh on AI sector investor confidence.
Throughout his Seoul visit, Huang conducted meetings with leadership from SK Group, Samsung, LG Group, Hyundai Motor Group, and Naver to address production scaling obligations and physical AI partnership opportunities.
NVIDIA has confirmed scheduled engagements with all principal South Korean technology and industrial corporations as it pursues worldwide Vera Rubin deployment expansion.


