Nvidia (NVDA) Selects Samsung and SK Hynix as Exclusive HBM4 Partners, Micron (MU) Left Out

Key Takeaways

Table of Contents

Nvidia has selected Samsung and SK Hynix as exclusive HBM4 memory providers for Vera Rubin, its next-generation AI accelerator
Micron is excluded from the Vera Rubin supply agreement, causing MU shares to decline 6.74%
Samsung successfully completed Nvidia’s HBM4 qualification at both 10 Gbps and 11 Gbps speeds; SK Hynix continues testing at 11 Gbps
SK Hynix is projected to deliver more than 50% of Nvidia’s HBM requirements in 2026; Samsung’s portion increases to 28%
Manufacturing is scheduled to commence in March, with Vera Rubin’s market debut planned for late 2026

Nvidia has designated Samsung and SK Hynix as the sole providers of sixth-generation high-bandwidth memory (HBM4) for the Vera Rubin AI accelerator platform, the Korea Economic Daily reports. Notably absent from this arrangement is Micron, which previously served as an important HBM partner.

The announcement triggered a 6.74% decline in Micron shares. Samsung’s Korea-listed stock fell 7.81%, while SK Hynix dropped 9.52%. Nvidia experienced a 3.01% pullback.

Micron Technology, Inc., MU

Vera Rubin represents Nvidia’s upcoming flagship AI platform, set to replace the current Blackwell architecture. The complete NVL72 rack setup combines 72 Rubin GPUs with 36 Vera CPUs and achieves 10x superior performance-per-watt compared to Blackwell.

Micron isn’t completely excluded from Nvidia’s roadmap. The company will provide HBM4 for Rubin CPX, a mid-range inference-oriented accelerator within the Rubin family. However, it won’t participate in the flagship Vera Rubin offering.

Samsung received approval after successfully meeting Nvidia’s stringent quality standards at 10 Gbps and 11 Gbps performance levels. SK Hynix continues validation efforts for the 11 Gbps specification but maintains its position as the largest HBM supplier globally.

SK Hynix Maintains Dominance, Samsung Expands Market Position

SK Hynix is anticipated to control approximately 50% of worldwide HBM production in 2026, a slight decrease from 59% in 2025. Samsung’s market share is expected to expand to 28%, rising from 20% the previous year.

SK Hynix is forecast to deliver over half of Nvidia’s combined HBM requirements — encompassing HBM3E — throughout 2026, and will likely lead HBM4 volume for Vera Rubin.

Both manufacturers are set to initiate HBM4 production this month. Vera Rubin remains on schedule for a second-half 2026 release.

Vera Rubin’s Target Applications

The Vera Rubin platform targets large-scale AI training and inference workloads, particularly the mixture-of-experts (MoE) architectures increasingly adopted in cutting-edge AI development.

Reported prospective customers include Microsoft, Amazon, Oracle, and Google. These cloud hyperscalers have been Nvidia’s primary customers in recent cycles.

The Vera Rubin NVL72 consumes twice the power of Blackwell while delivering significantly improved efficiency per watt, a critical factor for data center operators deploying these systems at massive scale.

HBM4 provides greater memory bandwidth than earlier generations, addressing one of the primary constraints in training and operating large-scale AI models.

Wall Street sentiment toward Nvidia remains optimistic. According to TipRanks, NVDA holds a Strong Buy consensus from 39 analysts, with one Hold rating. The average price target of $272.16 suggests approximately 53% potential upside from current trading levels.

Over the trailing 12 months, Nvidia stock has appreciated 66.2%, despite Monday’s decline following the supplier announcement.

Samsung and SK Hynix are preparing to launch HBM4 production in March 2026, with Vera Rubin systems expected to ship during the second half of that year.

✨ Limited Time Offer