Micron reportedly targets 2× HBM capacity by end-2026; Samsung, SK Hynix keep the scale edge - digitimes

Micron is reportedly planning to double its high-bandwidth memory (HBM) capacity by the end of 2026, targeting approximately 100,000 wafers per month, primarily to meet demand for Nvidia's Vera Rubin ... Read original →

Micron’s plan to double HBM capacity by end-2026, driven by demand for Nvidia’s Vera Rubin AI platform’s 12-layer HBM4, signals a shift toward large-scale adoption of high-bandwidth memory in AI computing. This move will catalyze upstream upgrades in wafer fabrication, photoresist, and packaging technologies, especially in 12-high architecture. From a compliance standpoint, geopolitical tensions and supply chain fragmentation increase operational risks, particularly in key manufacturing hubs like Taiwan, China and Hong Kong, China, where companies must weigh investment returns against security concerns. Samsung and SK Hynix, with their scale advantages, may respond with pricing or innovation strategies to maintain dominance. In the medium term, this expansion will accelerate AI chip performance, reshaping the compute landscape around memory-centric architectures. The next 12–24 months will likely see heightened demand for memory bandwidth, forcing the entire semiconductor ecosystem to align with evolving AI platform requirements.