Micron warns AI‑driven memory shortage will linger past 2027 as industry ramps up $200B investment

Micron CEO Sanjay Mehrotra told CNBC that the global memory‑chip shortage is not caused solely by the rapid rise of AI. Years of price pressure forced manufacturers to cut capital spending, leaving production capacity unable to meet the sudden surge in demand for high‑bandwidth DRAM and HBM used in AI servers. Micron’s gross margin fell to –7.3 % in fiscal 2023 and capital expenditures dropped from $12.1 bn to $7.7 bn, but the company is still investing roughly $200 bn in new capacity and research, including new fabs in Boise, Idaho, and Syracuse, New York. Mehrotra expects the shortage to persist until well after 2027 because building new factories and advanced‑node memory takes several years.

Industry analysts at SEMI project that 300 mm memory‑fab equipment spending will exceed $50 bn in 2026, a 29 % rise driven by AI‑related demand for DRAM, HBM and 3D NAND. Global capacity is forecast to reach 4.1 million wafers per month in 2026. At the same time, advanced semiconductor packaging is expanding, with larger 2.5 D and 3D packages, panel‑level packaging, and new glass interposer technologies aimed at integrating more compute and memory dies for AI and high‑performance computing.

The combined effect of AI‑driven demand, constrained supply, and rising investment underscores a pivotal shift in the memory market, with implications for device pricing, cloud‑service costs, and the broader tech ecosystem.