MLCCs: The Small Part Creating Big Infrastructure Risk

by Brad Gastwirth Global Head of Research and Market Intelligence

While GPUs, HBM, and substrates dominate the AI infrastructure narrative, large-case MLCCs (multilayer ceramic capacitors) are emerging as a critical, and often overlooked, constraint. Their availability could quietly derail server and rack deployment schedules, even when silicon is secured.

What’s Driving the Risk AI Servers Need More MLCCs Than Ever

  • AI servers now require 3,000–4,000 MLCCs per board, more than double the count in traditional servers.
  • High-speed computing environments demand high-capacitance, high-voltage MLCCs, especially near GPUs where power integrity is critical.

Demand Is Surging Across Sectors

  • AI infrastructure including servers, accelerators, and power systems is now a primary driver of MLCC demand.
  • EVs and industrial systems are also competing for the same large-case MLCC formats, intensifying pressure on supply.

Supply Chain Is Concentrated and Fragile

  • Murata, TDK, Taiyo Yuden, and Samsung Electro-Mechanics dominate production, with facilities concentrated in Japan, Korea, and China.
  • Despite strong demand, MLCC suppliers are not scaling large-case production fast enough, due to yield sensitivity and underinvestment.

Substitution Isn’t Simple

  • A single 1812 MLCC can replace 4–8 smaller capacitors. If forced to substitute with smaller sizes, boards may require 2–3x more units, increasing BOM complexity and tightening global supply.

Why Larger MLCCs Matter Larger MLCCs (1210, 1812+) offer several advantages that make them indispensable in AI infrastructure:

  • Higher Capacitance per Unit: Fewer components needed to meet power requirements.
  • Better Thermal and Voltage Handling: Ideal for high-power environments like GPUs and accelerators.
  • Reduced Parasitics: Fewer parts means lower inductance and resistance, improving signal and power stability.
  • Board Optimization: Can simplify layout by replacing clusters of smaller capacitors.
  • Greater Reliability: More robust under thermal and mechanical stress — important for data centers and industrial-grade systems.

Strategic Implications

  • A missing $0.20 MLCC can delay shipment of a $30,000 AI server.
  • Even with GPUs secured, hyperscaler deployment timelines are vulnerable to passive component shortages.
  • If substitution becomes widespread, global MLCC demand could spike by billions of units annually, straining capacity and logistics. Forward Indicators to Watch
  • Lead Times: Large-case MLCCs (1210/1812) trending toward 20–24 weeks.
  • Allocation Notices: Watch for rationing in high-voltage X7R/X6S grades.
  • Spot Market Activity: Pricing stability in broker channels despite macro softness.

Risk Score: 10/15 — High (Red Zone) Risk Thermometer

🔴 Critical (13–15): Allocation risk and shipment delays likely.

🟠 High (10–12): Current status — elevated risk.

🟡 Medium (5–9): Manageable with buffers and multi-sourcing.

🟢 Low (0–4): Stable supply and standard lead times.