PMIC supplier GMT sees upside in potential AI spending slowdown

The PMIC value chain is being repriced as AI capex enters a digestion phase. GMT's thesis is fundamentally a capacity-release argument: once data-center demand for 28-55nm mature-node PMICs shifts from crowding out to freeing up, consumer-electronics BOMs see immediate relief on power-delivery components. The deeper cascade matters more—when AI accelerators stop siphoning HBM and DDR5 supply, memory pricing normalizes, and given the 12-18-month PMIC design-in cycle, the order-book mix will visibly rotate by H1 2026. Competitively, TI and NXP retain structural moats in industrial and automotive PMIC, but Chinese players like SG Micro and Southchip are exploiting the AI-cooling window to accelerate mid-tier consumer penetration. Geopolitically, US export controls target advanced packaging and HBM—not the mature nodes where PMICs reside—creating a de facto compliance safe zone for Taiwan, China and mainland PMIC supply chains. The 12-24-month call: AI infrastructure transitions from arms-race to absorption. PMIC demand restructures rather than contracts. The real winners are not AI-PMIC specialists but second-tier suppliers with dual consumer-automotive exposure and sufficient capacity elasticity.