Murata to invest JPY1 trillion in expanding MLCC capacity across Asia plants

Murata's 1-trillion-yen commitment is less a capex announcement than a strategic repositioning: MLCCs are being reclassified from consumer-electronics consumables to the "power arteries" of AI infrastructure. When a single GPU crosses 700 W and rack density pushes toward 100 kW, board-level decoupling capacitor demand scales non-linearly. This is an architectural shift, not a cyclical bump. Upstream, BaTiO3 powder and ultra-thin nickel inner-electrode capacity will hit bottlenecks before finished-device output does. Downstream, server ODMs will be forced to redesign their power-distribution networks, making >100-layer, 01005-case MLCCs a de facto standard on AI accelerator cards. On competition, Samsung Electro-Mechanics and TDK will likely mirror this capex in H2 2025, but Chinese players—Fenghua, Sanhuan—already own the cost curve in mid-to-low tiers. Murata's deliberate "across Asia" framing, rather than naming a single country, is both a geopolitical hedge and a supply-chain-resilience signal to hyperscaler customers. The real risk: if AI capex growth decelerates in 2026, depreciation drag will compress ASPs back to 2023 levels, and second-tier exits will accelerate faster than consensus expects. The 12-to-24-month variable is not capacity—it is the deployment cadence of AI training clusters. Murata is betting that the "power wall" hits before the "compute wall." That thesis, for now, holds.