Published on: 2026-09-01
Updated on: 2026-09-01
Artificial intelligence infrastructure is usually discussed in terms of GPUs, memory chips, networking equipment and data centres. The newly launched Defiance AI Capacitors Leaders ETF (Cboe: CAPA) takes a narrower approach by targeting the passive electronic components that help those systems receive stable power and operate reliably.
Launched in August 2026, CAPA tracks companies involved in multilayer ceramic capacitors, conductive polymer capacitors and other advanced passive components used in AI servers and related infrastructure. The result is not another conventional semiconductor ETF. Instead, CAPA gives investors exposure to a concentrated group of mainly Asian component manufacturers positioned further down the AI hardware supply chain.
CAPA launched on August 26, 2026, tracking the BITA AI Capacitors Leaders Index with a 0.71% expense ratio.
The ETF targets capacitor and passive-component companies supporting AI infrastructure rather than semiconductor manufacturers themselves.
Its benchmark is highly concentrated, with the five largest index constituents accounting for roughly 92% as of August 28.
CAPA currently gets its capacitor-stock exposure primarily through a total-return swap on the BITA index rather than directly holding the ten index constituents.
CAPA is the Defiance AI Capacitors Leaders ETF, a U.S.-listed thematic exchange-traded fund designed to provide exposure to companies involved in the capacitor and passive-component supply chain supporting artificial intelligence infrastructure.
The fund seeks to track the performance, before fees and expenses, of the BITA AI Capacitors Leaders Index. Rather than focusing on AI software developers or semiconductor designers, the ETF targets companies that manufacture components used to condition power, regulate voltage, filter signals, suppress electrical noise and store energy inside servers, accelerators, networking systems and data centres.
That distinction is central to understanding CAPA.
A GPU may perform the computation, but it still depends on thousands of passive components positioned throughout a server board and power-delivery system. As AI accelerators draw more current and server power architectures become more demanding, Defiance’s investment thesis is that the capacitor content and technical requirements per system should continue increasing.
CAPA therefore attempts to capture a different part of the AI infrastructure buildout: the components supporting the chips rather than the chips themselves.
However, that does not make every CAPA constituent a pure-play AI capacitor business. The index methodology also permits companies with material involvement in the broader capacitor ecosystem, meaning investors are buying exposure to established electronic-component manufacturers whose revenues may extend well beyond artificial intelligence.
CAPA began trading on August 26, 2026, making it one of the newest thematic ETFs targeting a specialised part of the AI infrastructure supply chain.
Defiance’s launch materials list an expense ratio of 0.71%. At that rate, an investment of $10,000 would correspond to approximately $71 in annual fund expenses, assuming the fee remained unchanged.
Because the ETF is so new, investors do not yet have a meaningful operating history from which to assess how closely CAPA tracks its benchmark across different market environments.
There are already some early trading figures, however. As of August 28, Defiance reported $510,780 in net assets, 20,000 shares outstanding, a $25.54 NAV and a $26.50 closing market price, equivalent to a 3.76% premium to NAV. The fund page also displayed a median 30-day spread of 0.12%.
A few trading sessions are far too little data from which to judge CAPA’s normal liquidity or premium-and-discount behaviour. Still, the early premium is notable because many of the benchmark companies trade on Asian exchanges.
Differences between U.S. trading hours and those of the underlying foreign markets can contribute to ETF market prices temporarily diverging from NAV.
CAPA tracks the BITA AI Capacitors Leaders Index, a rules-based benchmark designed to identify publicly listed companies involved in advanced capacitors and passive electronic components used in AI infrastructure.
Eligible activities include areas such as:
multilayer ceramic capacitors, or MLCCs;
conductive polymer capacitors;
server-grade aluminium electrolytic capacitors;
polymer hybrid capacitors;
related passive components serving AI servers and data centres.
The methodology uses two routes for determining whether a company belongs in the theme.
A company can qualify by deriving at least 50% of its total revenue from eligible business segments. Alternatively, it can qualify by demonstrating material involvement in the ecosystem through factors including supply-chain importance, research and development, disclosed investment, strategic partnerships, government contracts or patent activity.
That second route is important because it makes the index broader than a strict pure-play screen.
Investors should therefore not assume that every CAPA constituent earns at least half of its revenue specifically from AI-related capacitors. The benchmark is designed to capture companies judged important to the theme as well as those meeting the explicit revenue threshold.
Companies must also satisfy minimum investability requirements, including a market capitalisation of at least $100 million, three-month average daily traded value of at least $100,000, and free float of at least 10%.
Current Defiance and Cboe materials describe the index as weighted by free-float market capitalisation, subject to a 20% maximum issuer weight, and rebalanced quarterly in January, April, July and October.
There is, however, an apparent inconsistency in the published documentation. An SEC summary prospectus dated August 24 states that the initial ten constituents would be equally weighted, while current issuer and exchange materials describe a free-float market-cap methodology with a 20% issuer cap. Published constituent weights also included companies above 20% even before CAPA began trading.
Investors should therefore check the latest index methodology and fund disclosures rather than assume that the phrase “20% cap” means every published constituent weight will always remain below 20%.
As of August 28, 2026, the BITA AI Capacitors Leaders Index contained ten companies:
| Company | Index Weight |
|---|---|
| TDK | 22.37% |
| Samsung Electro-Mechanics | 20.90% |
| Murata Manufacturing | 18.22% |
| Yageo | 15.56% |
| Kyocera | 14.92% |
| Taiyo Yuden | 4.20% |
| Walsin Technology | 1.69% |
| Maruwa | 1.69% |
| Samwha Capacitor | 0.34% |
| Nippon Chemi-Con | 0.11% |
Index holdings as of August 28, 2026. Weights and constituents can change.
The headline number is ten holdings, but the economic exposure is much narrower.
TDK, Samsung Electro-Mechanics, Murata, Yageo and Kyocera together represented roughly 92% of the index.
That makes CAPA a highly concentrated thematic fund rather than a broadly diversified electronics portfolio. Performance will be driven disproportionately by developments affecting those five companies.
There is also substantial geographic concentration.
Most of the benchmark is composed of companies headquartered in Japan, South Korea and Taiwan, reflecting where much of the advanced capacitor and passive-component manufacturing industry is located.
CAPA may therefore trade on a U.S. exchange, but investors are also exposed to foreign equity markets, currency movements, regional manufacturing cycles and differences in market hours.
These companies represent CAPA’s benchmark exposure, not necessarily the securities physically held by the ETF.
CAPA’s current portfolio looks very different from the ten-stock index above.
As of August 31, 2026, Defiance disclosed the following positions:
| CAPA Holding | Weight |
|---|---|
| BITA AI Capacitors Leaders Index total-return swap | 100.55% |
| First American Government Obligations Fund | 76.35% |
| Cash & other | -76.91% |
Fund holdings as of August 31, 2026. Holdings can change daily.
In other words, CAPA was obtaining essentially all of its capacitor-stock exposure through a total-return swap linked to the BITA AI Capacitors Leaders Index, rather than by directly owning TDK, Samsung Electro-Mechanics, Murata and the other benchmark constituents.
The percentages can initially appear unusual because derivative exposure is measured differently from ordinary stock ownership. A swap can provide economic exposure based on its notional value while the ETF separately holds cash-management or collateral positions, producing gross percentages above 100% and offsetting negative cash or other balances.
CAPA’s prospectus permits this structure. It allows the fund to gain index exposure through component securities, depositary receipts and derivatives including options and total-return swaps, and explicitly states that the index may be replicated synthetically.
That means there are two separate concepts:
Index constituents determine the companies whose performance drives CAPA’s benchmark.
Fund holdings show the securities, cash and derivative positions CAPA actually holds on a particular day.
This distinction is particularly important for CAPA because several benchmark constituents trade primarily on overseas exchanges. Synthetic replication can provide another way to obtain that exposure, although swaps also introduce risks including counterparty exposure, financing costs, liquidity risk and imperfect tracking.
Investors wanting to know exactly how CAPA is obtaining its exposure should therefore look at the fund’s current daily holdings rather than rely solely on the index constituent list.
CAPA and semiconductor ETFs can both benefit from expanding AI infrastructure spending, but they target different parts of the hardware chain.
Traditional semiconductor ETFs generally invest across chip designers, foundries, memory manufacturers and semiconductor-equipment companies. CAPA instead focuses on the components used to stabilise power, filter electrical signals and manage voltage around processors and other electronic systems.
Its thesis is that AI infrastructure could require more capacitors per server, higher-specification components and increasingly complex power-delivery systems.
The portfolio structure is also different. Semiconductor ETFs often spread exposure across dozens of liquid listed companies. CAPA’s benchmark contained only ten names as of August 28, with roughly 92% concentrated in five companies, while the fund itself was obtaining its benchmark exposure primarily through a total-return swap as of August 31.
A different position in the supply chain therefore does not automatically mean broad diversification. CAPA remains a concentrated thematic exposure that can still be affected by global electronics cycles, manufacturing conditions and changes in technology spending.
MLCC stands for multilayer ceramic capacitor.
An MLCC is a small electronic component constructed from alternating layers of ceramic dielectric material and conductive electrodes. Despite its size, it performs several essential functions inside electronic circuits.
MLCCs can help:
stabilise voltage;
filter electrical noise;
smooth fluctuations in power delivery;
temporarily store electrical energy;
decouple different parts of a circuit.
These functions become increasingly important around advanced processors.
AI accelerators operate at relatively low voltages while drawing large amounts of current. Their power requirements can also change extremely quickly as computational workloads rise or fall.
The voltage supplied to the processor nevertheless needs to remain stable.
Capacitors positioned close to the chip help absorb these rapid fluctuations and supply or store small bursts of energy when required.
There is independent evidence that AI infrastructure is already affecting the MLCC market. TrendForce reported in July 2026 that AI demand pushed monthly shipments from Murata, Samsung Electro-Mechanics and Taiyo Yuden to a five-year high. It also reported late-June book-to-bill ratios of 1.30, 1.31 and 1.25, respectively, their strongest levels since the pandemic.
Murata’s own published materials have likewise identified AI servers as an important IT-infrastructure growth opportunity and projected substantial growth in MLCC demand from AI-equipped servers through 2030.
According to industry reporting cited by Defiance, a server based on Nvidia’s GB300 platform can require roughly 30,000 MLCCs, while a full AI rack can consume hundreds of thousands of capacitors. Murata has also been cited as forecasting its MLCC shipments into AI servers to grow at approximately 30% annually through 2030, reaching more than three times their 2025 level.
These figures should be treated as industry estimates and manufacturer forecasts, not universal engineering specifications or guarantees of future demand.
The investment thesis is therefore not simply that AI servers use large numbers of inexpensive capacitors.
Specification matters as well as quantity.
As racks consume more power and electrical architectures move from traditional designs toward higher-voltage systems, manufacturers may need components capable of handling tougher reliability, capacitance, temperature and power requirements.
If that trend continues, AI could increase both the number and value of advanced capacitors used per system.
CAPA offers something meaningfully different from the large semiconductor and broad AI ETFs already available.
It provides a U.S.-listed route into a specialised passive-component supply chain dominated by companies in Japan, South Korea and Taiwan. The underlying argument is straightforward: increasingly powerful AI infrastructure requires increasingly sophisticated power delivery, and capacitors are one of the components that make that possible.
But the ETF’s narrow focus also creates its main risks.
Around 92% of the index was concentrated in five companies at the end of August 2026. Those businesses are not necessarily pure AI plays, meaning CAPA’s performance can still be influenced by consumer electronics, industrial demand, currency movements, component pricing and the wider technology hardware cycle.
Its implementation also deserves attention. As of August 31, CAPA was obtaining essentially all of its benchmark exposure through a total-return swap rather than directly owning the index’s ten constituent stocks. That adds another layer for investors to monitor alongside concentration, foreign-market exposure and tracking behaviour.
CAPA is therefore better understood not as another bet on AI processors, but as a concentrated investment in one of the component layers those processors depend on.
Whether that becomes a durable source of returns will ultimately depend on more than the number of AI servers being built. Investors will also need to watch capacitor content per system, pricing, manufacturing capacity, competitive conditions, CAPA’s replication structure and how much of the industry’s projected AI demand actually translates into earnings for the companies driving its benchmark.
