AI infrastructure supply shortages are extending from chips and electronic components to the equipment needed to manufacture semiconductors. Lead times have reached as much as 4.7 times normal levels, while South Korean equipment makers report that critical parts previously delivered in four months now take 10 months, six months or longer, and in one Japan-sourced laser-processing case up to 40 months. The bottlenecks threaten schedules for AI servers, data centers and semiconductor fabs, including Samsung Electronics and SK Hynix projects under construction in Pyeongtaek and Yongin. TrendForce found shortages in five of six tracked infrastructure categories, with GPUs the only exception at their 20–30-week equilibrium range. ABF substrates faced the tightest conditions at 48–56 weeks, while HDDs reached 50 weeks, DRAM 20 weeks, MLCCs 30 weeks and enterprise SSDs 16 weeks. The AI boom has also pushed manufacturers to redirect MLCC output toward premium server components, lifting prices despite weak smartphone and PC demand. TSMC has repeatedly raised its 2026 equipment procurement outlook, reinforcing pressure through the semiconductor supply chain. In China, a surge in cleanroom orders during the first half of 2026 points to stronger domestic semiconductor equipment demand in the second half, based on the typical five-to-nine-month lag between cleanroom construction and equipment installation. Upstream suppliers remain reluctant to expand capacity because the duration of the AI-driven semiconductor cycle is uncertain, leaving the shortage increasingly structural and raising the risk that global production infrastructure investment will be postponed.