Why AI ETFs Haven't Taken Off

Thematic ETFs are popular and for good reason. Unlike standard classification systems that bucket stocks into sectors and industries, themes are tied to trends and narratives that tend to be more intuitive and more compelling to many investors.

One of the hottest themes in recent memory is artificial intelligence. And yet, for all the excitement around AI, the amount of money in AI-themed ETFs is surprisingly small relative to the size of the trend—just $36 billion.

Why is that? Perhaps because many of the biggest beneficiaries of AI are available in funds investors already own.

The AI Value Chain

Start with Nvidia. It’s uncontroversial to call it the single biggest beneficiary of the AI boom. The company sells the chips and systems that power the data centers running AI workloads.

But Nvidia is also the largest stock in the broader market, carrying roughly an 8% weight in the Vanguard S&P 500 ETF (VOO). Investors who own a simple index fund already have significant Nvidia exposure.

Other obvious beneficiaries are the semiconductor companies in Nvidia’s supply chain. Taiwan Semiconductor fabricates Nvidia’s chips. ASML supplies the lithography equipment TSMC relies on. KLA and Applied Materials provide additional chipmaking tools. Micron and other memory manufacturers supply the high-bandwidth memory that goes into AI systems.

All of these companies are well represented in semiconductor ETFs like the VanEck Semiconductor ETF (SMH) and the iShares Semiconductor ETF (SOXX). In practice, those funds have become de facto AI ETFs.

The next layer consists of the hyperscale cloud companies, including Microsoft with Azure, Google with Google Cloud and Amazon with AWS. These firms are building out the infrastructure to deploy AI at scale. 

But they’re also massive holdings in the S&P 500, with Alphabet at roughly 5.4%, Microsoft at 5%, and Amazon at 3.4% of the S&P 500. Investors already own them through broad market funds like VOO or large-cap funds like the Invesco QQQ Trust (QQQ).

Above the infrastructure layer sit the model companies. Two of the biggest, Anthropic and OpenAI, are private and unavailable to public market investors. Google’s Gemini is another major player, but again, Alphabet is already a top holding in VOO and QQQ.

The application layer has been more complicated. Many traditional software companies that were supposed to benefit from embedding AI into their tools have instead faced pressure as investors question whether they are being disrupted. 

Model providers are increasingly rolling out their own agents, coding assistants, and productivity tools, raising concerns that incumbents could be squeezed. 

There have been pockets of acceleration, particularly at large companies like Meta and Alphabet, but those are megacap names already heavily represented in broad market funds.

The Performance Case

Returns make the point even clearer. Over the past five years, QQQ is up 94% and VOO is up 90%. Both have outperformed the Global X Artificial Intelligence & Technology ETF (AIQ), which gained 73%, and the Global X Robotics & Artificial Intelligence ETF (BOTZ), which gained just 16%.

Semiconductor ETFs have done even better. SOXX is up 159% and SMH is up 242% over that same five-year stretch.

The largest AI-themed ETF, the $9.2 billion iShares AI Innovation and Tech Active ETF (BAI) has performed reasonably well since its October 2024 inception, gaining 43% compared with 24% for QQQ and 20% for VOO. But it still lags the 64% return for SMH and the 57% return for SOXX over the same period. 

A look at BAI’s top holdings shows a familiar lineup of semiconductor and megacap technology names, many of which also dominate SMH, SOXX, QQQ and VOO.

Where the Money Has Gone

Given all of this, investors have rationally gravitated toward semiconductor ETFs or large-cap funds for their AI exposure. SMH is a $49 billion fund while SOXX has $23 billion. QQQ is a giant with more than $400 billion in assets, and VOO has $870 billion.

Some investors are also expressing the AI theme through adjacent bets, buying traditional sector ETFs like the Utilities Select Sector SPDR Fund (XLU) or the Industrial Select Sector SPDR Fund (XLI) to gain exposure to the power and infrastructure buildout that AI requires.

Even so, the AI trade has been driven primarily by semiconductors, with megacap tech accounting for much of the rest. There has been little need for a dedicated AI fund when the core beneficiaries already dominate the most widely held ETFs in the market.

That could change if the gains from AI spread beyond chips and infrastructure into a broader swath of industries. Until then, investors appear content to get their AI exposure through the funds they already own.