Intel increases equipment orders as server demand exceeds supply

Intel's capacity expansion is fundamentally a forced resuscitation of x86 architecture driven by AI inference workloads. The real chokepoint isn't EUV lithography—ASML capacity has long ceased to be the binding constraint—but ABF packaging substrate lead times stretching beyond 40 weeks, compounded by the unexpected Israel facility disruption, exposing a structural single-point-of-failure vulnerability in Intel's global footprint. Upstream, Applied Materials and Lam Research now carry order visibility into 2026, yet the 12-18 month equipment delivery cycle creates a hard lag before meaningful capacity materializes. Downstream, Unimicron (Taiwan, China) and Ibiden stand to capture the first wave of substrate demand, with utilization rates likely reclaiming the 95%+ range. On the competitive front, AMD's EPYC 9005 series and NVIDIA's Grace CPU are steadily eroding Intel's share among hyperscalers. This expansion is defensive, not offensive—Intel must first secure the inventory-availability baseline before the 18A narrative gains traction. The 12-24 month thesis: rigid demand from AI inference clusters for multi-core x86 will keep server chip ASPs firm, but if Intel fails to resolve packaging yield bottlenecks by 2026, the expansion dividend will be progressively diluted by structural substitution from AMD and ARM-based alternatives.