“Samsung’s HBM4 delivers industry‑leading performance, and feedback from customers has been very positive,” said Samsung Electronics’ Chief Technology Officer Jaehyuk Song, speaking to reporters on February 11 in Seoul ahead of his keynote at Semicon Korea 2026.
The firm’s sixth‑generation high‑bandwidth memory (HBM4) is set to begin shipments after the Lunar New Year holiday to AI semiconductor giant Nvidia, marking the first time Samsung will supply HBM4 to the US tech leader.
Song described the achievement as a return to Samsung’s “core technological strengths,” underscoring ongoing work to maintain leadership through successive generations. “We are preparing technologies so that we can remain at the forefront with next‑generation products such as HBM4E and HBM5,” he said.
Samsung adopted an aggressive development strategy for HBM4, combining 10‑nanometer sixth‑generation (1c) DRAM with a 4‑nanometer foundry process for the base logic die. This vertical integration enables the HBM4 chips to achieve data‑processing speeds of up to 11.7 gigabits per second (Gbps) — about 37 % above the JEDEC industry standard of 8 Gbps and roughly 22 % faster than the previous HBM3E generation (9.6 Gbps). Memory bandwidth per stack reaches up to 3 terabytes per second, about 2.4 times that of the prior generation.
Industry observers view the shipment to Nvidia — anticipated during the third week of February — as a potential turning point for Samsung’s leadership in the HBM market. Nvidia plans to integrate Samsung’s HBM4 into its next‑generation AI accelerator platform, Vera Rubin, which the company is expected to showcase at its GTC 2026 conference in March.
While Samsung has moved quickly to complete customer qualification testing, the broader HBM4 supplier landscape remains competitive. SK hynix is also poised to begin HBM4 shipments this month and is expected to account for a significant share of Nvidia’s vendor volume, according to reports.
Song highlighted Samsung’s vertically integrated manufacturing strength — with memory, foundry, and packaging capabilities under one roof — as a key differentiator. “Samsung offers an environment ideally suited to meet the product requirements of AI workloads, and this synergy is maximizing our technology competitiveness,” he said.
To support growing demand, Samsung is also expanding HBM4‑specific DRAM production, particularly at its Pyeongtaek P4 facility, increasing output to address the intensified demand from AI customers such as Nvidia and AMD.
As the HBM4 era unfolds, industry analysts are watching closely how Samsung and other memory makers position themselves for market share in high‑performance memory — a critical component of next‑generation AI infrastructure.
