[News] Samsung Reportedly Set to Begin Official HBM4 Shipments to NVIDIA and AMD in February

Samsung appears to be making solid headway in HBM4, with recent progress reinforcing industry expectations that the company could regain its competitiveness in next-generation memory. According to Hankyung, industry sources say Samsung is poised to begin official HBM4 shipments next month to major AI chip customers such as NVIDIA and AMD. This development follows Samsung’s successful completion of final qualification tests, prompting customers to place mass-production orders rather than request only sample volumes.

As SBS Biz notes, sources say Samsung Electronics has recently cleared final HBM4 qualification tests conducted by NVIDIA and AMD, and is set to begin mass production in February. The report indicates that Samsung’s HBM4, shipped in February, are expected to be delivered to NVIDIA and immediately used in performance demonstrations of the Rubin AI accelerator, which is set to debut at GTC 2026 in March.

Samsung Reportedly Achieves Industry’s Highest HBM4 Specification

SBS Biz also reports that Samsung’s HBM4 achieves a data rate of 11.7 Gb per second, exceeding the 10 Gb speed requirement set by NVIDIA and AMD—positioning it as the highest-performing specification in the industry. Sources further indicate that the product passed validation without any redesign, even after customers requested enhanced performance last year.

As Hankyung adds, Samsung is aiming for maximum performance by adopting sixth-generation (1c) DRAM on a 10nm-class process—placing it one generation ahead of competitors—and applying a 4nm foundry process for the logic die, putting it several generations ahead in control-core integration.

In addition, Sources cited by Global Economic News say Samsung Electronics has secured more lead time than competitors dependent on TSMC by procuring its own 4nm logic die. The report also notes that NVIDIA has moved Vera Rubin into full-scale production and has urged suppliers to accelerate their response.

Meanwhile, SBS Biz notes that full-scale HBM4 supply is expected to begin around June. Since HBM4 is directly integrated into AI accelerators like NVIDIA’s Rubin, shipment volumes will depend on customers’ final product mass-production schedules. With major customers still engaged in next-generation chip manufacturing at foundries, Samsung’s HBM4 output is likely to be adjusted in line with their actual ramp-up timelines and demand.

Looking ahead, Hankyung reports that Samsung plans to leverage its fully integrated HBM capabilities—from logic-die design to packaging—to strengthen its leadership in future offerings such as HBM4E (7th generation) and custom HBM solutions, reinforcing its position in the next phase of HBM development.

Read more

• [News] Samsung Reportedly Moves Custom HBM Logic Die to 2nm Foundry Process for the First Time
• [News] Samsung & SK hynix Earnings Showdown on 1/29: Profit, CAPEX, HBM4 in Focus

(Photo credit: Samsung)