HBM

Samsung is expected to more than double its production of HBM4 and HBM4E high-bandwidth memory DRAM next year, according to a report citing sources. The expansion targets surging demand for advanced memory used in AI accelerators, where Samsung competes with SK Hynix and Micron for supply deals with chip and data-center customers.

SK hynix is considering producing memory chips in the United States, according to the Korea Herald. The South Korean chipmaker, a leading supplier of HBM and DRAM to companies like Nvidia, has so far concentrated manufacturing in Korea and China. A US move would align with Washington's push to bring advanced semiconductor production onshore through incentives and tariffs, though the company has reportedly not made a final decision.

A shortage of memory chips is being driven by surging AI infrastructure demand colliding with limited wafer capacity and the industry's focus on high-bandwidth memory for accelerators. DRAM and NAND makers have shifted investment toward HBM, leaving conventional memory supply tight, and new manufacturing capacity takes years to bring online. Discussion centres on how long the squeeze will last and what it means for pricing across the tech industry.

Jensen Huang has reportedly committed to doubling chip supply, a move that would lock up roughly 37% of the world's high-bandwidth memory (HBM) output for 2026. The pledge underscores Nvidia's aggressive push to secure memory for its AI accelerators amid shortages, leaving rival chipmakers competing for the remaining HBM capacity from suppliers like SK Hynix, Samsung and Micron.

Micron is being discussed as a key beneficiary of the AI boom, with attention focused on its memory chips used in AI data centres. The discussion centres on investment insights into whether the company's HBM and DRAM business can keep growing as demand for AI hardware accelerates.