Memory semiconductor supercycle set to run through 2028 (blocksandfiles.com)

🤖 AI Summary
The memory semiconductor market is entering a significant supercycle expected to last until at least 2028, driven largely by the surging demand for high-bandwidth memory (HBM) needed for artificial intelligence (AI) training and inference. Companies like Micron and SK Hynix are experiencing a shift from the traditional model of oversupply and undersupply to a sustained period of tightness, with Micron predicting that the HBM market will grow from $35 billion in 2025 to around $100 billion by 2028. This prolonged demand has resulted in DRAM prices soaring by 171% year-over-year, as memory suppliers prioritize HBM production over conventional DRAM. The implications of this supercycle are significant for the AI and broader semiconductor community. As manufacturers struggle to expand capacity to meet extraordinary demand, NAND prices and delivery times are expected to remain elevated, forcing companies to adapt by implementing innovative storage solutions. Strategies include moving non-critical data away from solid-state drives (SSDs) and employing data reduction techniques. This structural crisis could reshape supply chain dynamics in the storage industry, suggesting that while the current memory constraints present challenges, they also spur advancements in data management and storage efficiency to cope with the increasing demands of AI workloads.
Loading comments...
loading comments...