Samsung to Apply 2nm GAA Process to HBM5 Base Die, Targeting 50%+ Speed Improvement

The Quick Take

• Samsung Electronics will apply its 2-nanometer GAA (gate-all-around) process to HBM5 memory base die
• The 2nm process is expected to deliver more than 50% speed improvement over HBM4E in the next-generation product
• Samsung's VP of Foundry Technology stated the move will cement technology leadership and reduce time-to-market through turnkey integration

Samsung Electronics will apply its cutting-edge 2-nanometer gate-all-around process technology to the base die of its 8th-generation High Bandwidth Memory product, HBM5, according to both Chosun Ilbo and Newsis. The HBM5 base die will use GAA architecture to deliver more than 50 percent speed improvement over HBM4E, the current generation, with enhanced power efficiency and higher TSV (through-silicon via) density. Ku Ja-heum, Samsung's Vice President of Foundry Technology Development, stated that the 2nm integration reinforces Samsung's technology leadership by enabling design optimization across memory and foundry divisions through a turnkey solution approach.

Samsung's decision to use 2nm on HBM base die is strategically significant because it positions Samsung Foundry's most advanced process at the heart of the AI memory supply chain. HBM is the critical bandwidth bottleneck for AI accelerators including Nvidia's H100 and B200, and any performance advantage in HBM5 over competitor products could determine which customers—primarily hyperscalers and AI chip designers—specify Samsung versus SK Hynix's HBM offerings. The turnkey model, where Samsung manages both memory DRAM and base die fabrication internally, also creates a manufacturing integration advantage that pure-play memory makers or outsourced assemblers cannot replicate.

Investors should watch Samsung Foundry's yield announcements for 2nm production, which is the critical execution challenge: advanced nodes have historically faced significant yield ramp challenges that delayed volume production and eroded competitive windows. The macro variable is the AI training and inference accelerator demand trajectory—sustained hyperscaler AI capex at current levels keeps HBM demand structurally elevated, validating Samsung's investment in cutting-edge HBM5 process integration. SK Hynix's competing HBM timeline announcements will set the competitive pace that determines whether Samsung's 2nm HBM5 strategy arrives first or plays catch-up.

Synthesized from 2 sources.