Samsung Electronics will commence mass production and shipment of its sixth-generation HBM4 high-bandwidth memory chips later this month for Nvidia graphics processing units powering the Vera Rubin AI accelerator platform, set to launch in the second half of 2026, with shipments starting as early as the week after the Lunar New Year holiday.
The HBM4 chips achieve data-processing speeds of up to 11.7 gigabits per second. This performance exceeds the JEDEC industry standard of 8 gigabits per second by 37 percent. Compared to the previous HBM3E generation, the speed improvement stands at 22 percent. Memory bandwidth per stack reaches up to 3 terabytes per second. This figure amounts to roughly 2.4 times the bandwidth of the HBM3E predecessor.
Samsung employs a vertically integrated manufacturing model for these chips. Unlike competitors that depend on external foundries like Taiwan Semiconductor Manufacturing Company for the base logic die, Samsung produces this component using its internal 4-nanometer foundry process. The company combines this with its 1c DRAM technology, which represents its 10-nanometer-class sixth-generation memory process.
An industry source stated to the Korea JoongAng Daily, “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.” This positions Samsung as the initial producer of these top-tier HBM4 chips on a mass scale.
Samsung gains a timing edge over SK Hynix, its main competitor. SK Hynix delayed its HBM4 mass-production from February to March or April 2026. Currently dominant in the HBM market, SK Hynix plans to continue HBM3E as its primary product through at least the first half of 2026. This decision stems from adjustments in Nvidia’s product strategy.
Samsung completed Nvidia’s quality certification process and obtained purchase orders. The company aligned its production schedule with Nvidia’s plans for the Vera Rubin launch. In Nvidia’s tests, Samsung’s HBM4 chips earned the highest evaluation scores for operating speed and power efficiency.
To address rising demand for AI memory, Samsung intends to expand HBM production capacity by about 50 percent by the end of 2026. The target reaches roughly 250,000 wafers per month, up from the current 170,000 wafers. At its Pyeongtaek Plant 4 facility, a new DRAM production line will contribute 100,000 to 120,000 additional wafers of monthly capacity. This expansion equates to an 18 percent increase in Samsung’s overall DRAM production capability.
