Taiyo Yuden Plans to Raise MLCC Annual Capacity Growth to 15%, Refuses to Hike Prices Amid Supply Tightness
N.R. Finch
Taiyo Yuden plans to lift MLCC annual capacity growth from 10% to 15%, yet explicitly refuses to follow the industry's price-hike wave — a long-horizon bet on volume over pricing as the AI-driven supply gap may last through 2028.
How many MLCCs does a single AI server need?
MLCCs — multi-layer ceramic capacitors, the most basic energy-storage components on virtually every circuit board — account for roughly 70% of Taiyo Yuden's total revenue.
A single AI server can require up to 28,000 MLCCs. Nvidia's next-gen Rubin architecture may need close to 12,000 per motherboard, nearly double the current level.
This means → AI data-center buildout is turning MLCCs from a low-value, high-volume commodity into a capacity-constrained strategic material.
The whole industry is raising prices — why won't Taiyo Yuden?
High-capacity MLCCs for AI servers have risen 15%–35%; some scarce part numbers have doubled on the spot market. Mid-to-low-capacity consumer and automotive MLCCs are up 6%–13%. Overall MLCC spot prices have climbed 20%–40%.
CEO Katsuya Sase's logic: capacitors are not memory chips. Historically, high standardization and multi-vendor interchangeability have kept MLCC prices on a long-term downward trend. The 2018 surge by Taiwanese suppliers was a historical exception, not a template.
In plain terms = when memory prices spike, customers have nowhere to go; when capacitor prices spike, customers can switch suppliers overnight. Aggressive pricing that backfires once the market loosens costs more in lost share than it earns in short-term margin.
Price increases already implemented cover rising raw-material costs such as silver — they are not shortage premiums.
What are the competitors doing?
Murata Manufacturing and Samsung Electro-Mechanics are also expanding, but long equipment lead times and production bottlenecks keep industry-wide capacity growth at an estimated 10%–15% per year.
Samsung Electro-Mechanics is taking a different path: developing packaging that integrates silicon capacitors with FC-BGA substrates. It has already landed a record single order worth ₩1.5 trillion.
This means → the top-tier race is forking. Taiyo Yuden bets on "scale + pricing restraint"; Samsung bets on "differentiated packaging." The two strategies will collide head-on during this super-cycle.
Are Chinese suppliers catching up?
Chaozhou Three-Circle has developed proprietary ceramic powders and improved multi-layer production processes, pushing high-capacity MLCCs into the Tesla and Nvidia supply chains.
This reflects a shift: China's domestic substitution in high-end passive components is no longer a concept — it has reached the approved-vendor lists of top-tier customers.
Japanese and Korean suppliers still dominate the high end, but Chinese challengers are adding a new competitive variable to this MLCC cycle.
Can Taiyo Yuden's gamble pay off?
The core thesis: use 15% annual capacity growth to capture the AI demand ramp, and pricing restraint to lock in long-term relationships with hyperscale data-center customers.
The supply-demand gap is expected to last through 2027–2028, giving the expansion strategy enough runway — but that also means excess capacity becomes a burden if demand falls short.
In plain terms = this is a "build first, harvest share later" long-horizon bet. The payoff hinges on whether AI data-center construction really sustains its pace through 2028, as the market currently expects.
Content is for reference only, not financial advice.