SEOUL, Jan 26, 2026, 22:17 KST
- Samsung plans to start making next-generation HBM4 memory chips next month and aims to supply Nvidia, a source said
- The move would help Samsung narrow the gap with SK Hynix in high-end AI memory
- Analysts say tight HBM supply across vendors could limit near-term share shifts
Samsung Electronics plans to start production of its next-generation high-bandwidth memory chips, or HBM4, next month and supply them to Nvidia, a person familiar with the matter told Reuters. Reuters
High-bandwidth memory, or HBM, is a stacked form of DRAM that sits close to AI processors to move data faster. Nvidia uses vast amounts of it for its AI accelerators, and supply has been tight as data center spending stays hot.
For Samsung, landing a bigger slot at Nvidia matters because it has been trying to catch up with cross-town rival SK Hynix, which has been a primary supplier of advanced memory chips used in Nvidia’s top AI systems. Samsung’s earlier HBM supply delays weighed on earnings and its share price, and investors have been watching for signs the company can ramp the next generation.
The source declined to give details such as how many chips Samsung plans to supply. A Samsung spokesperson declined to comment, while Nvidia was not immediately available for comment.
Bloomberg reported Samsung has entered the final qualification phase with Nvidia after providing initial samples to the U.S. chipmaker in September, and is preparing for mass production of HBM4 in February. The company will be ready to ship soon, though the timing is not yet clear, the report said. Bloomberg
Samsung shares rose in early trading in Seoul, while SK Hynix shares fell. The moves came as investors bet Samsung could finally join the tight circle supplying the most advanced memory chips that Nvidia pairs with its flagship AI processors.
A separate report by South Korea’s Korea Economic Daily said Samsung passed HBM4 qualification tests for Nvidia and Advanced Micro Devices and will start shipping to Nvidia next month, citing chip industry sources.
SK Hynix has also been building out capacity. It has said it completed HBM supply talks with major customers for next year, and plans to begin deploying silicon wafers next month into a new fab, M15X, in Cheongju, South Korea, to produce HBM chips, without saying whether HBM4 would be part of initial production.
Nvidia CEO Jensen Huang said earlier this month that the company’s next-generation chips, the Vera Rubin platform, is in “full production,” as it prepares to launch the chips later this year to be paired with HBM4.
Barron’s flagged the knock-on effect for Micron Technology, another major HBM supplier. William Blair analyst Sebastien Naji said Micron’s HBM production for 2026 is already sold out and expects the company to keep HBM market share in the low-20% range through 2027, arguing that tight supply across Samsung, SK Hynix and Micron may limit a rapid reshuffle in who wins the biggest orders. Barron’s
But the hinge is still Nvidia’s testing. Final qualification can turn on yields — the share of usable chips coming off a line — and on performance under Nvidia’s workloads. Any slip in Samsung’s ramp, or a longer validation cycle, would leave SK Hynix in the driver’s seat a bit longer, and keep Samsung’s volumes smaller than the market is starting to price in.
Both Samsung and SK Hynix report fourth-quarter results on Thursday, Jan 29, when investors expect more detail on HBM4 orders, production plans and how fast the next supply wave can reach Nvidia’s systems.
