Global Semiconductors: eSSD Demand to Rise Sharply in 2027E on AI Continual Learning

Continual learning and AI inference are set to drive a sharp increase in eSSD demand, which we forecast to grow +52.9% YoY in 2027E. We expect QLC demand to rise significantly, driven by near-GPU storage adoption and increasing demand from AI data centers in China. We believe strong enterprise AI demand should offset softer consumer NAND demand. As memory makers prioritize DRAM/HBM capacity expansion over NAND, NAND supply-demand conditions should tighten materially, resulting in deficits of -6.1%/-5.5% in 27E/28E. We expect both training and inference demand to continue growing rapidly from 27E onwards, as continual learning technologies become increasingly sophisticated. This should lift eSSD demand for AI inference, as continual learning requires AI systems to continuously incorporate new information while retaining previously learned data. We also expect rising demand for advanced solutions such as XL Flash and HBF, alongside storage-complementary technologies such as CXL. In addit