Micron Commits $24B to Expand NAND Capacity in Singapore to Ease AI-Driven Shortages

Micron has disclosed a major capital program to expand NAND wafer manufacturing at its Singapore complex, committing about $24 billion to add roughly 700,000 square feet of controlled cleanroom space with production slated for the second half of 2028. The company frames the investment as a response to accelerating demand for memory from artificial intelligence workloads and hyperscale data centers, which have tightened NAND inventories. This NAND expansion sits alongside a previously announced $7 billion high-bandwidth memory (HBM) packaging facility in the same cluster; Micron expects the HBM line to begin materially easing HBM tightness in 2027. Executives say they will pace incremental capacity additions to market signals, reflecting lessons from prior memory cycles and the long lead times of wafer fab projects. The program emphasizes automation, AI-driven process control and robotics to both raise yield potential and constrain operating headcount growth. Micron projects roughly 1,600 fabrication engineering and operations roles tied to the NAND expansion, with about 1,400 roles already associated with the HBM packaging plant — together creating roughly 3,000 technical and manufacturing positions in Singapore. The move strengthens Singapore’s standing in the semiconductor supply chain and attracted visible support from the city-state’s economic development agency. The announcement arrives as other large semiconductor players are accelerating verified capex and pursuing supplier qualification for next-generation HBM, a pattern that validates demand but also sharpens competition for tools, materials and hyperscaler commitments. Key execution risks remain: equipment lead times, wafer and package qualification cycles, yield ramping and the need for sustained hyperscaler purchasing to justify the outlay. If Micron executes on schedule and demand endures, the added NAND throughput should materially relieve flash tightness by the end of the decade; if not, the company could face elevated fixed costs amid volatile memory pricing. Overall, the investment is an aggressive, long-lead strategic bet that pairs near-term HBM packaging capacity with longer-lead NAND wafer growth to capture AI-driven memory demand while attempting to avoid past overbuild pitfalls.