Yageo plans 10-15% capacity expansion in 2027 as AI demand pushes utilization above 95%

Yageo's 10-15% capacity expansion is not incremental scaling—it is a structural response to AI compute pushing passive components toward physical density limits. Once rack power density crosses 30kW, MLCC per-board counts scale exponentially, and tantalum capacitors regain pricing power in high-voltage filtering. The 4pp QoQ jump in AI revenue to roughly 20% signals a durable shift from training to inference workloads, not a one-time procurement spike. Upstream, BaTiO3 ceramic powder and tantalum ore face tightening constraints. Tantalum mining remains concentrated in DRC and Mozambique, while China's expanding critical-mineral export controls add friction for Yageo, headquartered in Taiwan, China. Under EU Conflict Minerals Regulation audits, compliance costs within capex could climb from a 3-5% historical norm to 8% or above. Competitively, Murata will likely ring-fence incremental capacity for automotive-grade high-capacitance MLCCs, while Samsung Electro-Mechanics competes on price in AI-server tantalum. Yageo's real adversary is the chip architecture cycle: a 2027 shift to 48V DC bus topologies could invalidate current MLCC specs, compressing the expansion window to roughly 12 months. Over the next 12-24 months, expect a widening scissors gap between AI-driven premium pricing and consumer-electronics destocking. If AI capex decelerates in H2 2027, 95%+ utilization flips into a price war, with MLCC ASPs potentially falling 15-20%.