#AI股持续上涨还有哪些投资机会 AI large model training and inference are driving massive computing power demand. Upstream hardware is the foundation for industry expansion. In the past, the market focused more on GPU and AI server complete-system leaders; but as capacity building accelerates, the value of short-board links in the supply chain has gradually become more prominent—what the industry calls the “chamber-of-wood effect”: the maximum capacity of the entire industry chain is determined by the most supply-constrained link.
Large-scale model training requires massive data read/write; high-bandwidth memory (HBM) is a core supporting component for GPUs, and supply has remained structurally tight for the long term. As AI agents and multimodal models are deployed, storage-side demand for inference continues to grow. Storage chip and storage module companies have room for sustained volume expansion.
Large-scale model training requires massive data read/write; high-bandwidth memory (HBM) is a core supporting component for GPUs, and supply has remained structurally tight for the long term. As AI agents and multimodal models are deployed, storage-side demand for inference continues to grow. Storage chip and storage module companies have room for sustained volume expansion.