The next AI hardware bottleneck isn't memory. It's a tiny capacitor whose price already tripled.
Inventory levels for key products have generally fallen to 30 days or less.
Distributors that had built inventories in advance have raised prices by an average of 20–25%, while spot-market prices have climbed to two to three times their previous levels.
Why a capacitor is chip news
This part is smaller than a fingernail. It is called a capacitor. It stores a small burst of electricity and releases it to keep power steady inside a circuit. A single server board holds thousands of them, and the ones built by stacking ceramic layers are called MLCCs. AI servers draw far more power and tolerate far less noise than a regular PC, so they need many more MLCCs, and better ones.
What actually happened
TrendForce data cited by Jukan shows June shipments from Murata, Samsung Electro-Mechanics and Taiyo Yuden hit their highest level in five years. AI demand is what pushed their production lines there.
The peak during the worst MLCC shortage on record, in 2018, was 1.25. All three makers are now at or above that line. The "highest in five years" claim checks out against the data.
2026-07-06 · TrendForce
AI demand for high-end MLCCs pushes Japan and Korea suppliers' order backlogs to post-pandemic highs (TrendForce, 2026-07-06)Prices moved with them. Distributors who had stocked up ahead of time raised prices 20 to 25%, and spot prices climbed as much as three times their earlier level. Inventory on key parts has fallen to 30 days or less.
Why the shortage hit now
The cause is not demand alone. It is allocation. Murata, Samsung Electro-Mechanics and Taiyo Yuden are shifting lines that used to make consumer-grade parts (X5R) toward the higher AI-grade spec (X6S, X7R). A factory cannot double its output overnight, so a line that gains capacity on one side gives it up on the other. The parts actually running short are mid-capacity capacitors (1 to 22 microfarads) used in ordinary appliances and cars.
Who fills the gap the big three leave
This is the part Jukan's post does not mention. As the big three chase AI orders, Taiwan's Yageo and Walsin Technology are stepping into the consumer-grade orders they leave behind.
AI servers trigger an MLCC cannibalization wave, Taiwanese makers stand to benefit (BigGo Finance, 2026-07-06)A shortage at the top three becomes an order surge for the smaller Taiwanese makers. The same pattern played out in the 2018 MLCC crunch, when Taiwanese suppliers gained share.
What settles this
The question comes down to one thing. Is this a structural bottleneck that rides the AI server boom for several quarters, or a temporary supply knot that unwinds once inventory rebuilds. If Murata, Samsung Electro-Mechanics and Taiyo Yuden's book-to-bill ratios stay above 1.25 through Q3 and Q4, the structural read holds. If the inventory that fell below 30 days returns to normal by year-end, this was a knot, not a wall. While memory chip prices swing through headlines, a much smaller and cheaper part is quietly setting the pace of AI hardware production.
- AI server demand pushed June MLCC shipments from Murata, Samsung Electro-Mechanics and Taiyo Yuden to a five-year high.
- Spot prices climbed as much as three times over, distributor prices rose 20-25%, and key inventory fell below 30 days.
- The consumer-grade orders the big three left behind are now flowing to Taiwan's Yageo and Walsin Technology.
What happened next
Awaiting gradingTo be scored on whether Murata, Samsung Electro-Mechanics and Taiyo Yuden's book-to-bill ratios stay above 1.25 through Q3-Q4, and whether inventory that fell below 30 days returns to normal by year-end.
This author's record
- BullWhy did chip stocks keep falling even after a string of earnings beats?
- BullASML slid nearly 10% in two days on China's DUV scare, but the case behind it w…
- BullThe inside story of how Samsung is lifting DRAM output 15% at Hwaseong without…