Report: Samsung’s HBM4/HBM4E production may double next year, with product share rising from 40% to 80%

Samsung Electronics is making a major bet on next-generation high-bandwidth memory, fully pivoting toward high value-added products in the AI memory chip race.

According to Seoul Economic Daily, Samsung Electronics plans to increase the outsourced cleaning volume of "glass substrates"—an indispensable material in the HBM manufacturing process—from 20,000 pieces per month this year to 50,000 pieces per month next year, a 2.5-fold increase. Based on this, industry insiders expect the actual output of HBM4 and HBM4E to at least double next year compared to this year. At the same time, Samsung's total monthly HBM wafer input is expected to rise from about 180,000 this year to around 250,000 next year, an increase of nearly 40%.

Behind this expansion in output is Samsung's strategic focus on higher-generation products. With the gradual ramp-up of HBM4E mass production, the share of the HBM4 series in overall HBM shipments is expected to soar from about 40% this year to around 80% next year. This structural shift means that Samsung will allocate more capacity resources to higher-margin products.

Glass substrates are a key consumable in the HBM manufacturing process. During the grinding and thinning of DRAM wafers and the drilling of holes, the glass substrate is attached to the back of the wafer to prevent warping or cracking.

Since HBM requires stacking multiple layers of DRAM chips within a limited thickness, the importance of wafer thinning techniques and warpage control processes increases with the number of stacked layers. The HBM4 and HBM4E products, which Samsung is focusing on expanding, both center on 12-layer and above high-stack products, demanding even stricter requirements for glass substrate technology.

According to Seoul Economic Daily, Samsung's monthly demand for glass substrates has risen from 10,000 pieces last year to 20,000 this year, and will further increase to 50,000 next year. Industry analysts note that even considering the possibility of reusing glass substrates after cleaning and variations in actual consumption due to process inputs and yield, the 2.5-fold surge in demand strongly points to expectations that production of the HBM4 series will more than double.

In terms of mass production progress, Samsung began shipping HBM4 in volume in February this year. This product uses sixth-generation (1c) DRAM at the 10nm level and a base chip made with a 4nm process. In May of this year, Samsung also provided samples of the HBM4E 12-layer version to clients such as Nvidia.

Industry forecasts suggest that as HBM4E enters full-scale mass production next year, the share of the HBM4 series in Samsung's overall HBM shipments will climb sharply from about 40% this year to roughly 80%. One industry insider stated, "While expanding HBM output, Samsung appears to be placing high value-added HBM4 products at the core."

This product mix transformation will directly enhance the overall profitability of Samsung's HBM business and reinforce its competitive position in the AI accelerator memory market.

This round of Samsung's expansion is giving an obvious boost to its upstream supply chain. Outsourced cleaning volume for glass substrates will increase from 10,000 pieces per month to 50,000 within two years—a fivefold jump—prompting related suppliers' order volumes to grow accordingly.

From an industry perspective, the ongoing increase in HBM stack layers—from the early 8 and 12 layers to even higher—has made wafer thinning and support processes an increasingly prominent bottleneck in the manufacturing flow. The rapid growth in demand for glass substrates is not only a direct reflection of Samsung's capacity expansion but also underscores the intensified material consumption brought by the increased complexity of high-stack HBM processes.