Capacitor supply-chain bottlenecks are now driving a fresh round of Wall Street speculation around AI hardware, with investment managers racing to package the shortage into exchange-traded funds. The people actually building and powering the data centers get the bottlenecks. The fund managers get the fees.
Who Gets Turned Into a Trade
CNBC said there are roughly two dozen broad-based AI ETFs, but the latest rush is moving deeper into the guts of the machine. Niche funds are now targeting memory, through the Roundhill Memory ETF, or DRAM, and photonics companies tied to AI data-center optical networking innovation. Another growing bet centers on capacitors, the components that help stabilize the intense power demands in data centers. That’s the new frontier for financial extraction: not just AI itself, but every choke point inside it.
Global X, which launched its broad AI ETF, Global X AI & Technology, or AIQ, in 2018, is planning to launch a capacitor-focused ETF within a few weeks, according to Pedro Palandrani, Global X head of product research and development, speaking during CNBC's "ETF Edge" podcast earlier this week. CNBC said the company has a prospectus on file with the SEC for the Global X MLCC & Electronic Components ETF and is currently in a quiet period until the ETF launches. Palandrani dressed the pitch in the language of client management and market chatter, saying, "I always say these are very relatable concerns. Whether the theme is AI or geopolitics and its impact on defense tech, Global X is working with many financial advisors using these ETFs as a way to have conversations with clients on investing opportunities."
He added, "Most of these client conversations are happening on these thematic categories, how the portfolios are positioned to capitalize on those things that may be headlines grabbers. Investors recognize that very well." He also said, "Targeted exposures are top of mind. We will continue to see those targeted exposures within AI." And then the blunt part: "Where we see significant opportunity is offering targeted exposure to many of the bottlenecks within AI."
The Bottleneck Becomes the Product
CNBC said the ETF market focus on multilayer ceramic capacitor manufacturers began last month with the launch in August of two portfolios: the more pure-play Defiance AI Capacitors Leaders ETF, or CAPA, and the broader Tema MLCC & PowerSemi ETF, or PSOX. CAPA is a bet on 10 companies based in Asia that dominate the capacitor market: TDK, Samsung Electro-Mechanics, Murata, Kyocera, Yageo, Taiyo Yuden, Maruwa, Walsin, Samwha Capacitor and Nippon Chemi-Con.
Tema said in its fund overview, "Every AI accelerator, server board, and hyperscale data center depends on capacitors and passive components to condition power, regulate voltage, filter signals, suppress noise, and store energy." Murata said that "a single AI server built on Nvidia's GB300 platform can require roughly 30,000 multilayer ceramic capacitors, with a full rack consuming hundreds of thousands of units." That’s the scale of the machine. And that’s the scale of the market feeding on it.
CNBC said Murata, the largest MLCC manufacturer, has forecast to investors that MLCC shipments will grow at approximately 30% annually through 2030, tripling 2025 levels, and that its stock price has risen over 115% year-to-date. The company’s growth projections and stock surge sit right beside the shortages themselves, a neat arrangement for anyone selling exposure to scarcity.
The Global X MLCC ETF will track the Akros MLCC & Electronic Components Index, which seeks to hold up to 15 companies that are among the largest MLCC manufacturers worldwide or have at least 50% revenue exposure to the component category, according to its prospectus. Tema's PSOX ETF targets some of those same capacitor makers, plus companies that focus on other aspects of power regulation within the AI market, including U.S.-based Bloom Energy and Monolithic Power Systems. CNBC said both thematic trades "represent emerging AI bottlenecks," according to Tema's fund materials.
Fees, Hype, and the Old Cycle
Roundhill Investments, whose DRAM ETF has been a breakout star this year with close to $25 billion in assets, is planning to get in on the trade as well through its CCML ETF, also currently in registration with the SEC. CNBC said the MLCC ETFs are off to a much slower start than DRAM and other AI trades in terms of attracting investor interest and asset gathering, with CAPA reporting assets of roughly $3.5 million and PSOX not yet at the $1 million mark. The fees don’t shrink for the small players. CAPA's annual expense ratio is 0.75% and PSOX's is 0.71%, both high relative to core equity holdings, though in line with what other thematic ETFs charge.
CNBC said recent press reports on the capacitor market bottleneck have referred to it as "the rice" of the AI buildout story. Taiyo Yuden’s CEO, whose shares are up over 150% this year, told Bloomberg in June that the company was facing "scary" levels of demand, especially as more advanced chips like Nvidia's Rubin hit the market. The demand is scary. The packaging of that demand into tradable products is apparently thrilling.
Todd Sohn, chief ETF strategist at Baird Strategas, called the market "a thematic tidal wave of ETFs, particularly centered around AI" and said, "These thematic booms tend to come in cycles." Nate Geraci, president of NovaDius Wealth Management, said, "It's a tale as old as time in asset management, where fund issuers rush to package and sell whatever is capturing investors' attention. Nothing is hotter right now than slicing and dicing the AI trade." He added, "If investors are going to dabble in these products, it's critical they look under the hood to understand exactly what they own and how it fits with the rest of their portfolio."
Sohn said the "slicing and dicing of these segments" marks a change from the historical role of ETFs as broadly diversified index-tracking vehicles. He said, "You just had to have asset allocation correct and let the market work over time ... limited stock knowledge was needed." He also said these ETFs can serve a real need for traders who want efficient access to companies powering a certain segment of the AI boom: "quick and efficient access to practically any market or theme." He warned investors to avoid fund overlap, adding, "If you own one broad AI ETF and overlay it with one of the niche funds, you may have more exposure to certain bellwethers than you need."
The whole setup is a familiar one. A supply chain strain appears. Asset managers rush in. The bottleneck becomes a product. The product becomes a fee stream. And the people at the bottom of the chain, the ones dealing with the actual power demands of the AI buildout, remain the raw material for somebody else’s portfolio strategy.