Morgan Stanley: MLCC Market Enters Structural Divergence — Murata Upgraded to Top Pick, TDK Taiyo Yuden Downgraded
Taylor Wilson
Morgan Stanley named Murata Manufacturing its top MLCC pick with a target price raised to ¥12,500, while downgrading Taiyo Yuden from EW to UW — opposite ratings on the same sector, because the call is that AI-server-driven MLCC demand is not a rising-tide beta but a tiered contest over high-end part-number delivery.
How hungry are AI servers for MLCCs?
A single AI server motherboard carries 15,000 to 25,000 MLCCs — roughly 10× a conventional server.
The reason: GPUs and CPUs switch at nanosecond speeds, creating huge transient current demands. MLCCs — multilayer ceramic capacitors, tiny components placed next to chips to discharge fast, suppress noise, and stabilize voltage — must sit right beside the die to do their job.
This means → AI servers don't just use "a bit more" MLCCs; they raise demand by an order of magnitude — and that is the starting point for the entire sector being re-priced.
How big is this market, and how fast is it growing?
Global MLCC shipment value in 2025: roughly $14.67 billion, up 10.3% year-on-year. The ten-year CAGR (2015–2025) was about 6.5% — a mature cyclical product.
Morgan Stanley projects $24.25 billion by 2028 (2025–2028 CAGR of 18.2%), rising further to $33.71 billion by 2031.
The critical number: high-value, small-size, high-capacitance MLCCs demanded by AI and data centers could see a three-year demand CAGR near 100%. In plain terms = the whole industry grows at a moderate clip, but the most valuable slice is roughly doubling every year.
Why is Murata the top pick?
Share dominance: Murata holds about 40.8% of global MLCC shipment value in 2025; SEMCO about 22.5%, Taiyo Yuden about 11.3%.
Technology accumulated long before the AI cycle: Murata mass-produced 1005-size 1µF in 2000, then 0402-size 1µF and 1005-size 22µF in 2020 — miniaturization and high-capacitance capability built over two decades, not rushed for this cycle.
The three key specs AI servers need most (1608-size 100µF, 1005-size 47µF, 0603-size 10µF) — only Murata can supply them at scale today. By the time rivals achieve stable mass production, Murata will likely have moved to even smaller, higher-capacitance next-gen parts.
This means → Murata's moat is not just share; it is a one-generation lead in mass production — competitors are always chasing the previous spec.
Why was Taiyo Yuden downgraded?
Taiyo Yuden is not performing badly — the report expects its MLCC sales and profits to keep growing, with utilization rates improving.
The problem: the stock already priced in "benefiting at the same tier as Murata," and the report judges that expectation is wrong.
The gaps: global share slipped from 12.7% in 2020 to about 11.3% in 2025; capacity additions are only about 5% for the current fiscal year (partly due to a prior capex freeze); and critically, Taiyo Yuden has not yet achieved full mass production of the key high-capacitance MLCCs needed for AI server accelerator cards.
In plain terms = Taiyo Yuden isn't weak — the market just priced it as if it were Murata, and its actual delivery capability doesn't support that valuation.
Is this a pricing cycle or a structural shift?
The report's view: even if a given MLCC spec sees short-term price hikes, over a year-plus horizon, its price will most likely drift lower.
But if the share of high-value products keeps rising, overall average selling prices still climb. This means → the core logic of this MLCC cycle is not "same spec, higher price" but "selling pricier specs."
The former is cyclical (what goes up comes down); the latter is structural margin expansion — and the competitive barriers each requires are entirely different. This reflects the deeper logic behind Morgan Stanley's opposite ratings: companies that can reliably deliver high-end part numbers capture structural profits; those that only ride the volume wave capture cyclical beta.
How does this show up in the financials?
Murata's AI/DC-related MLCC as a share of total MLCC sales: roughly 10%–15% in F3/26 (fiscal year ending March 2026), expected to rise to 20%–25% in F3/27.
If the product-mix improvement materializes, Murata's ROE rises from 8.8% in F3/26 to 12.2% in F3/27 and 16.7% in F3/28.
Taiyo Yuden's AI/DC MLCC share is projected to go from 5%–10% to about 15% — a noticeably smaller step-up than Murata's. This reflects two companies on the same track drawing visibly different profit-curve slopes.
Content is for reference only, not financial advice.