Published on: June 12, 2026
One AI Rack Now Eats 440,000 MLCCs: How the Passive Squeeze Is Pushing General-Purpose Capacitors Into the Spot and Secondary Channel
A single GB300 rack consumes nearly 9x the MLCCs of a traditional server, and the Big Three are routing capacity to high-margin AI parts. The real story isn't the price hike — it's general-purpose and industrial MLCC allocation getting crowded out, with orders spilling into the secondary and spot channel. Here's what's loosening, what's tightening, and what to stock.
Start with one number.
A traditional server uses roughly 50,000 MLCCs. NVIDIA's GB300 AI rack needs around 440,000. Close to 9x.
This isn't a pricing headline. It's an allocation-logic headline.
MLCC capacity is fixed. Murata, Taiyo Yuden, and Samsung Electro-Mechanics are now prioritizing high-capacitance, high-voltage parts for AI servers — higher ASP, fatter margins. General-purpose, consumer, and mid/low-end automotive parts drop down the production queue.
The result:
- High-cap AI parts: lead times stretched from 8–12 weeks to 16–24 weeks, with order restrictions starting on some specs.
- General / industrial parts: makers aren't short on them, but allocation is being squeezed out, and orders spill into the secondary and spot channel.
- Spot pricing: already up 15–20%.
The price hikes are a continuation — Murata, Taiyo Yuden, and Yageo rolled out 15–35% increases from April 1. That's background, not this week's news. What to watch this week is the allocation spillover.
For anyone trading parts, both the opportunity and the trap sit on that spillover line.
What's loosening: Not the parts — the orders. Makers are handing general-purpose volume to the secondary and Taiwanese second-tier lines (Walsin, Yageo secondary). That puts two kinds of stock on the market: idle general-purpose MLCC bled out by end-customer cancellations, and alternate parts from secondary channels. The former is exactly what you want to buy cheap and match against urgent shortfalls.
What's tightening: High-cap AI server parts and high-reliability automotive grades. Don't touch these without maker allocation — you can't secure stable supply, spot pricing is inflated, and one mixed date code and you're stuck holding. Unless you have a firm end-customer order that matches the exact part, date code, and quantity, don't stock it.
Action list for spot buyers:
- If you're sitting on idle industrial/consumer general-purpose MLCC, don't dump it cheap right now. Spillover will force some end demand into the spot channel — a matching part number becomes a premium order.
- On AI high-cap inquiries, verify date code and maker traceability first. Don't take unknown-origin spot — this is where counterfeits and refurbished parts surface fastest.
- For industrial, energy-storage, and EV-charger customers carrying heavy general-purpose MLCC in their BOMs, lock Q3 supply early. Maker lead times are still stretching; locking late means price hike plus shortage at once.
- Don't bet on the high-reliability automotive spread. The allocation isn't yours — that table belongs to makers and tier-one distributors.
The MLCC line is the same script as the HBM memory-capacity grab of the past two years — high-end parts vacuum up capacity, general-purpose parts take the collateral hit. The difference: passives are low-ASP, high-volume, and fragmented across thousands of part numbers, which actually favors the middleman. Cancelled end orders leave more idle stock floating, and urgent shortfalls are scattered — perfect for matching and capturing the spread.
Whoever holds the matching parts, and can verify date code and quantity fast, is who catches this spillover.