AI needs a reboot

The AI trade is starting to short-circuit. After years of red-hot momentum, the rally is hitting a few glitches:

• Nvidia shares fell 6.5% last quarter, marking two straight quarterly losses and its longest losing streak since 2022.
• Memory companies Micron and SanDisk slid last week after Google unveiled TurboQuant, a tool that could cut AI memory needs by up to six times—spooking investors betting on endless demand.
• Super Micro Computer is grappling with legal overhang after its cofounder was caught smuggling chips to China last month (he pleaded not guilty yesterday). The company says it’s complying with export laws, but the stock has already dropped nearly 25% since the news broke.
• Chipmakers Broadcom, AMD, and TSMC are getting dragged by geopolitics, as the Iran conflict threatens higher energy costs across Asia—raising production expenses and supply chain risks—while roughly a third of global helium supply, critical for chipmaking, has been disrupted.

The good news

Put it all together, and it’s safe to say that tech is struggling—a sharp reversal after years of market dominance. The shift reflects growing skepticism around massive AI spending colliding with geopolitical risks and supply chain constraints. But the pullback has flushed out positioning and reset valuations, potentially setting the stage for outperformance.

The Magnificent 7 as a group now trades below 25x forward earnings, well off its 33x peak in October and under its 10-year average of 29x. Meanwhile, Mag 7 fundamentals remain intact, with profits projected to grow 19% in 2026, outpacing the broader market’s 14%, according to Bloomberg.

For Nvidia in particular, the near-term weakness may follow a familiar pattern. Historically, after streaks of two or more monthly declines, the stock has averaged a roughly 12% gain in the following month, suggesting momentum could quickly flip.

Zooming in on memory stocks, some analysts see the recent selloff as misplaced. Mizuho points to a version of the Jevons paradox: efficiency gains from tools like Google’s TurboQuant could actually accelerate AI adoption, ultimately boosting demand for memory over time.

As for geopolitical risks, UBS analysts say the helium disruption is manageable for now. Existing inventories should cushion the impact, and if the conflict resolves within two to three months, the effect on chip production is likely limited.

The outlook for tech stocks remains murky, but one thing is clear: the selloff has made valuations look a lot less stretched. With prices reset and earnings still holding up, the recent weakness may be less a warning sign and more an opportunity.—SY