Micron anticipates that the chip shortage will start to "improve gradually" by 2028 - WN Hub

Micron CEO Sanjay Mehrotra projected that the global shortage of RAM and NAND, driven by AI advancements, will persist until 2027, with gradual improvement expected in 2028. Despite ongoing supply con... Read original →

The AI compute arms race has triggered structural—not cyclical—shortages in DRAM and NAND. Bottlenecks in DDR5/HBM yields, delayed EUV tool deliveries, and U.S.-Japan-Netherlands export controls have elongated capacity ramp timelines. Consequently, Micron, Samsung, and SK Hynix are locking supply with hyperscalers like Microsoft and NVIDIA, relegating consumer devices to secondary allocation status—Valve’s $1,049 Steam Machine is merely a symptom. Compounding risks, U.S. CHIPS Act ‘guardrails’ restrict China-focused expansions, while advanced memory output from Taiwan, China and mainland China remains insufficient. Even with aggressive CapEx, talent gaps and fab construction lags will sustain tight supply through late 2027. Critically, if CXL-based memory pooling fails to scale, the ‘memory wall’—not compute—will become AI’s true bottleneck as bandwidth demand doubles every 18 months.