NEW YORK, March 30, 2026, 11:13 EDT
- VGT, CHAT and AIQ slipped Monday morning, with investors reconsidering their approach to the AI theme.
- New market chatter is turning focus from just the megacap platforms, pushing names in chips, memory, fiber, and connector supply up the watchlist.
- Low-cost, broad technology funds are pulling further ahead of their smaller, narrowly focused AI counterparts.
Investors seeking to stay in the artificial intelligence trade are showing a clearer preference: broad exchange-traded funds versus more targeted plays in chip, memory, and networking stocks that power the sector. By Monday morning, Vanguard Information Technology ETF was down close to 1.0%. Shares of Roundhill Generative AI & Technology ETF slid about 1.1%, while Global X Artificial Intelligence & Technology ETF dropped 0.8%.
This shift is catching attention right now: money is moving back into U.S. equities, but tech isn’t seeing the same love. In the week to March 25, U.S. equity funds raked in $37.24 billion. Tech sector funds, on the other hand, lost $1.45 billion. Nvidia’s forward price-to-earnings ratio? It’s dropped to its lowest since early 2019, with investors weighing just how quickly the big cloud players can convert massive AI outlays into real returns.
The question is sparking different takes among fund managers. Giuseppe Perrone, president and managing partner at Varenne Capital Partners, pointed out Monday that “performance dispersion” is spreading out across the big tech names. He laid out an AI “value-chain approach” as a way to broaden exposure—moving up the chain to suppliers like TSMC, Micron, Corning, and Amphenol, or further down to other segments. PWMNet
If it’s broad and inexpensive you’re after, Vanguard’s VGT still stands out as the straightforward pick. The ETF, which tracks a U.S. information-technology benchmark, carries a 0.09% annual fee. Vanguard announced last week it will split the shares 8-for-1 on April 21, cutting the share price but leaving the underlying portfolio intact.
CHAT’s on the opposite side of that trade. According to Roundhill, the fund—which debuted in May 2023—comes with a 0.75% fee. One recent analysis tagged it as a high-conviction play on the generative-AI stack. As of the end of 2025, Roundhill’s top five positions in CHAT included Alphabet, Nvidia, Microsoft, Meta Platforms, and SK Hynix.
Passive players aren’t standing by. Global X’s AIQ, following an artificial intelligence and big-data index, reported $7.12 billion in assets as of March 27. Its biggest bets: SK Hynix, Samsung Electronics, TSMC, Nvidia, and Micron—names clustered where BlackRock’s Jay Jacobs put it, “where the revenues are right now,” referring to the AI infrastructure surge earlier this year. Global X ETFs
Amplify’s AIVC ETF takes things up a notch, following Bloomberg’s AI Value Chain Index and splitting assets evenly among 45 stocks. The fund charges a 0.59% fee. That setup means investors get exposure to semiconductors, cloud software, and hardware suppliers—without piling into just a couple of big platform names.
Concentration remains a factor. According to Yahoo Finance, VGT’s three largest holdings are Nvidia at 18.07%, Apple at 15.84%, and Microsoft at 10.39%. Today, Reuters noted that investors continue to question when Microsoft, Alphabet, and Amazon will see enough returns from AI to make their capital outlays worthwhile. Glenmede’s Michael Reynolds, meanwhile, flagged that valuations for AI-exposed stocks are “getting a bit rich.” Yahoo Finance
Momentum in infrastructure hasn’t faded. Reuters pointed to a wave of hefty AI, cloud, and chip agreements—Nvidia, Meta, OpenAI, and several others all in the mix. But Bridgewater’s Greg Jensen flagged a shift: the frenzy, he said, is in a “more dangerous phase” now, with hard construction costs surging and a bigger role for outside money. Reuters
Wall Street isn’t debating AI’s investability anymore. Now it’s about pinpointing the sweet spot: broad, low-cost U.S. tech plays like VGT, more focused AI picks with CHAT, or funds targeting behind-the-scenes essentials—think foundries, memory, fiber, and connectors that power these models.
