Taiyo Yuden and Murata dominate the market for high-spec capacitors used in AI servers as soaring demand drives long-term contracts, capacity expansion and mounting price pressure.
Demand for high-performance MLCCs, or multilayer ceramic capacitors, is surging as AI data center buildouts lift component needs far beyond those of conventional servers. Taiyo Yuden President Katsuya Sase said order volumes for AI server parts were "scary" and that supply pressure was "extremely intense," underscoring how Japan's Taiyo Yuden and Murata Manufacturing have emerged as a key bottleneck in the market for premium MLCCs. A single AI server can use up to about 28,000 MLCCs, roughly 13 times a standard server and about 30 times a smartphone, because GPUs require dense clusters of capacitors near the processor to smooth sharp voltage swings. As chip performance rises, that requirement is increasing further: Nvidia's previous-generation GB200 used 6,500 MLCCs on one board, while the next-generation Rubin platform expected later this year is projected to need 12,000. TrendForce said AMD's next-generation MI450 platform also shifted toward MLCCs during validation, raising the count of certain MLCCs on the board to 10,544 from 1,440. Murata holds roughly 40% of the overall MLCC market, followed by Samsung Electro-Mechanics at around 20% and Taiyo Yuden at about 10%, but Bloomberg said Taiyo Yuden and Murata account for most global supply of high-performance MLCCs for AI servers. The tightening market is pushing customers toward three- to five-year long-term supply agreements and increasingly toward NCNR terms, while suppliers expand capacity. TrendForce projects global AI server shipments will grow 28.3% year over year in 2025 versus 12.8% for the broader server market, and book-to-bill ratios of 1.31 for Samsung Electro-Mechanics, 1.30 for Murata and 1.25 for Taiyo Yuden suggest orders continue to outpace shipments.