Is 5N high-purity dysprosium oxide the real bottleneck in the AI-MLCC supply chain?

Is 5N high-purity dysprosium oxide the real bottleneck in the AI-MLCC supply chain?
The AI server boom has exposed a critical dependency: high-capacity MLCC capacitors require 5N (99.999%) dysprosium oxide as a dielectric additive, with no mature substitute. Industry estimates suggest AI-MLCC alone could add 1,500 tonnes of dysprosium demand by 2027, nearly 40% of global supply of roughly 3,500 tonnes. Currently, only one Chinese company—Shenghe Resources' subsidiary Jiangyin Jiahua—achieves stable 5N production at about 15 tonnes annualized capacity. Japan's Shin-Etsu has suspended accepting new orders for dysprosium-containing high-end magnets, while Murata issued 10-40% price hikes on AI-server MLCCs effective July 2026. China's export halt to Japan since December 2025 has forced downstream ceramic powder makers to seek domestic alternatives. The purity gap between China's 5N capability and Western producers' 99.9% output creates a two-tier market where premium electronic-grade material commands significant price premiums.