Four US MLCC ETFs Launch as AI Power Needs Surge

The US market has seen a rapid proliferation of exchange-traded funds focused on multilayer ceramic capacitors, or MLCCs, with four new products listing within the past month. This concentration of capital reflects a strategic pivot toward power components as artificial intelligence data centers consume more electricity, shifting investment attention away from graphics processing units and memory chips.

According to Seoul Economic Daily, these funds include Themes ETFs' PSOX, Defiance's CAPA, Roundhill Investments' CCML, and Global X's MLCC. Three are listed on the Cboe BZX Exchange, while CCML trades on Nasdaq. The instruments vary slightly in scope, with PSOX and MLCC focusing on broad electronic components, CAPA targeting AI-specific capacitors, and CCML including printed circuit boards.

Samsung Electro-Mechanics holds top weightings

Samsung Electro-Mechanics emerges as the largest single holding across these new vehicles, carrying a 21.55% weighting in CAPA’s underlying index. The company also holds approximately 20% of Global X’s MLCC fund, roughly 17% in CCML, and around 9% in PSOX. Murata Manufacturing, the other dominant player in the sector, is included in all four products, confirming the duopoly structure of the high-end supply chain.

Market research firm TrendForce indicates that Murata leads the segment for MLCCs used in AI servers with a 45% share, while Samsung Electro-Mechanics holds 40%. This dominance stems from the requirement for compact, high-capacity components that stabilize voltage fluctuations. The surge in fund creation marks the first time Samsung Electro-Mechanics has been positioned as a core holding under a dedicated MLCC banner, rather than just a general electronics play.

GPU upgrades drive component count growth

The underlying driver for this demand is the increasing power consumption of high-performance GPUs. As processor capabilities rise, the need for reliable current supply and noise reduction grows, directly lifting the volume of high-capacity MLCCs required per unit. Kiwoom Securities estimates that the number of MLCCs per GPU will climb from 200 in Nvidia’s H100 to 500 in the B200, 1,500 in the GB200, and approximately 5,000 in the next-generation Rubin architecture.

This technological shift is projected to expand the market for AI server MLCCs from $1.3 billion last year to $5.8 billion by 2030. Asset managers are packaging this Asian supply chain as a standalone theme because the gains from expanding AI investment are concentrated in these specific component manufacturers. The financial logic rests on the direct correlation between GPU performance and passive component density.

Supply constraints extend lead times

Supply bottlenecks are creating a structural opportunity for incumbents. Large-volume supply of high-end MLCCs is effectively limited to Murata and Samsung Electro-Mechanics, as customer certification takes two to three years and new production lines require 12 to 24 months to reach mass production. Capacity additions at Samsung’s Philippines plant and Murata’s Izumo facility are not expected to have a full effect until 2027.

Kiwoom Securities analyst Kim Seung-hyuk notes that while industry capacity grows at 10% to 15% annually, AI server demand is estimated to rise more than 80% per year. Consequently, lead times, which typically range from six to eight weeks, have increased by at least 20 weeks. This supply-demand imbalance is expected to persist into next year, reinforcing the investment case for these specialized component providers.