Samsung Electronics will start mass production and shipment of its sixth-generation HBM4 memory chips later this month. Shipments could begin as early as next week after the Lunar New Year holiday. The chips target Nvidia graphics processing units and will support Nvidia’s Vera Rubin AI accelerator platform, set for launch in the second half of 2026.
HBM4 chips achieve data-processing speeds of up to 11.7 gigabits per second. This exceeds the JEDEC standard of 8 Gbps by 37 percent and surpasses the previous HBM3E generation by 22 percent. Memory bandwidth per stack reaches up to 3 terabytes per second, about 2.4 times higher than HBM3E.
Samsung uses a vertically integrated manufacturing process. It produces the base logic die with its internal 4-nanometer foundry and pairs it with 1c DRAM, its 10-nanometer-class sixth-generation memory technology. Industry sources note Samsung’s advantages: “Samsung, which has the world’s largest production capacity and the broadest product lineup, has demonstrated a recovery in its technological competitiveness by becoming the first to mass‑produce the highest‑performing HBM4,” an industry source told the Korea JoongAng Daily.
This positions Samsung ahead of rival SK Hynix. SK Hynix delayed its HBM4 mass production from February to March or April 2026. It plans to rely on HBM3E as its main product through at least the first half of 2026, influenced by changes in Nvidia’s product strategy.
Samsung completed Nvidia’s quality certification process and secured purchase orders. Its production schedule matches Nvidia’s Vera Rubin timeline. Samsung’s HBM4 chips earned the highest scores in Nvidia tests for operating speed and power efficiency.
To address rising AI memory demand, Samsung aims to boost HBM production capacity by 50 percent by the end of 2026. It targets 250,000 wafers per month, up from the current 170,000. A new DRAM production line at Pyeongtaek Plant 4 will add 100,000 to 120,000 wafers monthly, increasing overall DRAM production capability by 18 percent.
