Samsung is following a dynamic production strategy for its HBM solutions, with HBM4 entering the focus to regain AI memory momentum.

According to Chosun, Samsung is now utilizing half of its HBM production capacity on HBM4 to chase the AI memory market. Products with weaker demand have been reduced so the production can ramp up for the trending versions.

Samsung first shipped an HBM4 memory chip in February 2026. The company has recently unveiled HBM4E with further improvements. However, demand for HBM4 is currently trending due to its broad application.

Industry insiders revealed that the Korean tech giant has assigned about 75,000 of its 150,000 monthly HBM DRAM wafers to HBM4 production.

The remaining 75,000 wafers were allocated to the 12-layer HBM3E, whereas production of the 8-layer HBM3E has been temporarily halted.

Samsung has a limited share in Nvidia’s HBM3E supply chain, which is used in the Blackwell AI accelerator. Meanwhile, the HBM4 is the main force behind Nvidia’s next-generation Rubin.

Apart from  Nvidia, Samsung may have been getting increasing demand from major cloud companies like Google, Amazon, and Microsoft, which are developing in-house AI accelerators to reduce reliance on NVIDIA GPUs.

Riding the AI memory frenzy, SK Hynix became Korea’s most valuable tech company. It surpassed Samsung Electronics, breaking a record of over 25 years of dominance.

Read more: A 25-year Samsung record ends as SK Hynix rides the HBM success

The post Samsung gives HBM4 priority as it works to regain AI memory momentum appeared first on Sammy Fans.