The AI trade has been remarkably straightforward for the past two years. Buy an AI ETF, get exposure to the theme in one click, and watch the gains roll in. But according to Jack Fu, CEO of Draco Evolution, that easy ride may be coming to an end.

After billions poured into AI-focused exchange-traded funds, the broad exposure that once amplified returns is now masking a different kind of problem: concentration risk. Many investors think they're buying diversified AI exposure, Fu told MarketDash, but what they're actually getting is "the same few big tech stocks in different wrappers."

When Concentration Works Until It Doesn't

Here's what happened. Over the past two years, U.S.-listed technology and thematic ETFs attracted massive inflows as investors chased the AI narrative without needing to pick individual winners. Most of these funds are heavily weighted toward a handful of mega-cap technology companies, and that concentration worked beautifully during the hype phase.

Fu acknowledges the logic behind this setup. These companies have the cash, scale, and computing power to stay ahead. They're often the first to turn AI spending into actual revenue. When enthusiasm for AI was surging, having your fund packed with these names was exactly what you wanted.

But concentration cuts both ways. When a few stocks dominate a fund, any stumble from earnings misses, regulatory pressure, or valuation concerns can hammer returns across the board.

Take a look at the holdings. The Global X Artificial Intelligence & Technology ETF (AIQ), Roundhill Generative AI & Technology ETF (CHAT), and Dan Ives Wedbush AI Revolution ETF (IVES) all have heavy exposure to the "Magnificent 7" stocks. That means lots of Nvidia Corp. (NVDA), Alphabet (GOOG) (GOOGL), and Microsoft (MSFT), alongside chip stocks like Taiwan Semiconductor Manufacturing Co. (TSM) and Micron Technology (MU).

Draco's AI ETF (DRAI) takes a different approach with a more diversified, multi-asset structure. Its major holdings include the First American Funds Inc X Government Obligations Fund (FGXXX) and other debt funds, as well as leveraged equity ETFs like ProShares UltraPro QQQ (TQQQ) and Direxion Daily S&P 500 Bull 3x Shares (SPXL). Over the past year, it's up more than 30%, according to data from Benzinga Pro.

Structure Suddenly Matters

Fu believes we're entering a phase where how these funds are built will matter more than what they're called. Many AI ETFs simply track companies linked to AI without adjusting for market conditions or managing risk dynamically.

That worked fine when everything moved in the same direction. But as stock prices diverge and market conditions become less predictable, static AI ETFs are likely to face more strain, Fu says.

Over the past year, fixed income exposure in mixed-asset AI funds played a stabilizing role, Fu explained. It helped portfolios stay invested during volatile periods rather than forcing investors to reduce risk at exactly the wrong time.

Looking ahead, Fu expects attention and capital to shift toward networking, power, and grid equipment sectors, along with companies that use AI to drive "measurable productivity." In other words, the market will start caring more about who's actually making money from AI, not just who's talking about it.

"AI is still a powerful long-term trend," Fu said. But managing risk around that trend will "increasingly determine outcomes."

This shift could push investors toward AI ETFs that adjust their exposure rather than holding every stock with an AI label slapped on it. Simply owning an "AI ETF" may no longer be enough as investors become more discerning.

DRAI offers one example of that flexibility. Unlike most AI ETFs, it's not equity-only. During April's tariff headlines and the tech selloff late last year, the fund reduced equity exposure and shifted toward defensive assets like Treasuries, bonds, gold, and the U.S. dollar. That helped cushion some of the damage from broader market swings.

The One Number That Matters Most

If there's a single signal that could quickly change the outlook for AI ETFs, Fu points to spending plans from the largest cloud and technology companies.

As long as these companies keep investing heavily, the broader AI ecosystem stays supported. But if spending slows for several quarters, expectations could reset in a hurry.

"The money going into chips and data centers is still massive," Fu said. "But the market will care more about return on that spend, not just the headlines."

The stakes are high. Big Tech is ramping up AI investment at an unprecedented scale, with the "Magnificent 7" expected to pour nearly $400 billion into AI infrastructure this year alone.

For now, the spending continues. But Fu's message is clear: the next phase of AI investing won't reward broad exposure as blindly as the last one did. Investors who understand the difference between AI hype and AI structure will likely come out ahead.