Samsung plans to begin shipping next-generation HBM4 memory to Nvidia for use with its AI accelerator chips in February, according to multiple reports, after the company faced supply delays last year.
South Korean newspaper Korea Economic Daily said on Monday that Samsung had passed HBM4 qualification tests for Nvidia and AMD and would begin shipping the memory chips to the two companies in February, citing unnamed chip industry sources.
Reuters also reported similar plans, citing unnamed people.
AI chips
The reports did not indicate details such as how many chips Samsung planned to supply to Nvidia.
Samsung lost ground to competitor SK Hynix last year after its plans for high-bandwidth memory, or HBM, were hit by supply chain delays.
The memory is a critical part of the high-end AI chips that have become one of the tech industry’s most in-demand resources due to a boom in construction of AI-focused data centres.
As a result, memory chips have seen worldwide shortages, leading to a price spike for some consumer electronics.
SK Hynix said earlier this month that it plans to speed up the opening of a new factory by three months and will begin operating another new plant in February, as it works to meet demand.
‘Humongous demand’
Sungsoo Ryu, chief executive of SK Hynix America, told Reuters the company planned to open the first factory in a new chip facility in Yongin, 25 miles south of Seoul, three months early in February 2027 to help meet chip demand.
The company also plans to begin deploying silicon wafers next month to a new fabrication plant called M15X in Cheongju, South Korea, to produce HBM chips.
Ryu said the company was seeing “a lot of tremendous and humongous demand” for memory.
