Investors have too many questions about the AI buildout, and not enough answers
The Nasdaq-100 is on the verge of a correction as hot parts of the AI trade weigh on the market
After helping to drive major indexes such as the S&P 500 and Nasdaq Composite to record highs during the second quarter, semiconductor stocks took another beating on Tuesday.
Chip stocks have endured large selloffs before - the DeepSeek-inspired selloff in January 2025, which prompted a major wipeout for Nvidia shares (NVDA), was one notable example. But as semis have continued to move lower in July, this latest challenge is proving to be even more enduring. The PHLX Semiconductor Index SOX, a popular index of U.S.-traded semiconductor names, entered bear-market territory earlier this month, and has continued to move lower since. As of Tuesday, the index was off 24.5% from its late June record high.
The pressure from semis, coupled with continuing weakness in shares of some megacap tech giants like Microsoft $(MSFT)$ and Alphabet $(GOOGL)$ $(GOOG)$, has pushed the Nasdaq-100 NDX to the edge of correction territory. The index was down 9.4% from its early June peak, according to FactSet data.
Some major chip names that aren't in the index have fallen even harder. Sandisk $(SNDK)$ has seen its value fall by more than half since its late-June peak. As has often been the case lately, the latest bout of semiconductor volatility started overnight in Asia, where shares of South Korean memory giants Samsung (KR:005930) and SK Hynix (KR:000660) $(SKHY)$got clobbered, leading to another round of trading halts. Fears about Chinese competition were at least partly to blame.
When the semis trade started to wobble late last month, many analysts on Wall Street initially hoped the pain would be short-lived. With companies set to report another round of blockbuster earnings, many figured it was just a matter of time before another round of beats-and-raises from the hyperscalers would jerk shares of AI infrastructure plays back into the green.
However, that isn't how things have been playing out - at least so far. Nobody denies that demand for memory chips, GPUs and other essential components is expected to remain robust. But strong reports from Alphabet and Celestica $(CLS)$ have failed to quiet investors' concerns about the eventual return on all of this AI-related investment. Optimistic talk from CEOs including Nvidia's Jensen Huang and Amazon's Andy Jassy haven't moved the needle all that much.
"There are too many questions and not enough answers related to the AI buildout. The ROI from hyperscalers' investments and the scale of AI value creation are still open questions," wrote Dennis DeBusschere, chief market strategist at 22V Research, in commentary shared with MarketWatch.
During the second quarter, traders piled into leveraged products like ETFs to try and cash in on the red-hot rally in semis. Now, those positions are being washed out. But whether this is simply a positioning reset, or the start of a bigger pullback, remains to be seen.
"The question is, outside of a positioning cleanup... what gets the AI trade working again? Because it feels very unclear if great [numbers] are even going to be rewarded," wrote Jeffrey Favuzza, an equities trader at Jefferies, in commentary provided to MarketWatch.
The answer will ultimately hinge on the so-called hyperscalers financing much of the AI buildout, not only with cash on hand, but increasingly by issuing debt and equity. As the chart below shows, the decline in hyperscalers' forward free cash flow has been mostly offset by a surge in expected cash flows for a basket of semiconductor stocks.
Creeping doubts about the hyperscalers' return on investment have started to manifest in other ways. Widening spreads on credit default swaps tied to debt issued by Nvidia and the coterie of hyperscalers - a group that includes Alphabet, Amazon, Meta $(META)$, Microsoft and Oracle $(ORCL)$- have driven fears about how the rising cost of capital, or myriad other potential roadblocks, might disrupt the AI trade.
"There's a lot going on," said John Velis, BNY's Americas macro strategist.
There's been concern about reports from China on its ability to make advanced semiconductor chips. There also have been more questions about the cost of capital for the AI buildout in the U.S., as MarketWatch has reported.
"One of the things that has been on my mind," Velis said, is the argument that AI is going to reduce costs, lower inflation and allow interest rates to come down. When demand for capital is growing, with highly rated corporate-bond issuance likely to hit a record in 2026 and more tech IPOs coming out, that demand for capital can also pressure interest rates higher. "The cost of capital goes up," Velis said.
The Nasdaq Composite COMP fell Tuesday, leaving it 8.2% off its record close in early June, according to Dow Jones Market Data. The S&P 500 SPX gained 0.2%. The Dow Jones Industrial Average DJIA rose 1%, leaving it 0.6% from its early July record.
-Joseph Adinolfi -Joy Wiltermuth
Comments