Micron CEO warns memory shortage will worsen through 2028 as AI demand surges

Micron CEO Sanjay Mehrotra warned during the company's Q4 earnings call that memory availability will become increasingly constrained over the next two years. "We expect memory and storage supply demand conditions to be much tighter in fiscal 2027 and 2028 than they were in 2026," Mehrotra said, adding that "demand exceeding supply" is expected in both years.

The tightening stems largely from AI workload growth. Mehrotra cited rising demand for AI agents, larger models, and expanded context windows as primary drivers. Samsung issued a similar warning this week, noting that high-bandwidth memory (HBM) demand is surging and will claim an additional 10 percent of Samsung's manufacturing capacity next year. Since fab capacity is finite and new fabs take years to construct, that shift diverts resources from standard DRAM production.

Building new manufacturing capacity is the core constraint. "Clean rooms take a very long time to build," Mehrotra explained. "Even after they are built and first wafer output begins, production ramps up only gradually."

The implications reach beyond memory pricing. Graphics card manufacturers have two new GPU generations launching within this window (Vera Rubin and RDNA 5), alongside next-generation consoles and new CPUs from Intel and AMD. Sustained memory price increases could influence product launch timing and manufacturer suggested retail prices across the industry.

Not all industry voices agree on the shortage's severity. Acer's CEO reportedly called it "impossible for the shortage to last until 2030" and suggested some messaging around scarcity is designed to maintain margins and share prices. Michael Burry, known for predicting the 2008 financial crisis, has also questioned the durability of the shortage, citing the Acer CEO's position and accelerating timelines for supply stabilization.