Micron pauses work on $4.5 billion Taiwan plant, Bloomberg News reports - 1470 & 100.3 WMBD

Micron's halt on its $4.5B Taiwan, China fab is not a budget trim—it is a structural retreat from the HBM race. The line almost certainly targets HBM3E and next-gen DRAM nodes. Pausing it removes a critical capacity increment from the 2026 global HBM supply pool, precisely when NVIDIA's next-gen AI accelerators demand higher stack counts. Samsung and SK Hynix now hold an unchallenged pricing window in AI memory. On compliance, Micron as a US entity operating in Taiwan, China sits in an export-control gray zone. The pause reads as preemptive hedging—Washington's tightening review of semiconductor investment in the region, combined with Beijing's expanding countermeasure toolkit against US memory chips, makes a $4.5B sunk cost far cheaper than a production line caught in a regulatory chokepoint. This is risk pricing, not retreat. Competitively, Samsung's Pingshan and SK Hynix's Cheongju HBM expansions will not slow; they will accelerate long-term customer lock-ins. If Micron cannot restart within 12 months, its AI-server DRAM share slides below 15%, and HBM market share could dip under 10%. 12–24 month outlook: the industry enters an asymmetric "two giants, one laggard" phase. Micron will likely redirect capex to Idaho, New York, or Hiroshima, but yield-ramp cycles run 18+ months. That vacuum is the profit-harvesting window for the Korean duopoly.