AI Stock Selloff Wipes Out $1 Trillion as Rate Fears Grip Tech

The AI boom just hit a speed bump, and it’s a big one. Over $1 trillion evaporated from the market caps of the world’s 20 most valuable AI and semiconductor stocks this week alone, as investors suddenly remembered that building massive data centers actually costs money, and interest rates just went up.

Nvidia fell 1.8% in premarket trading Thursday, while Micron dropped 4.4% and AMD slid 3.5%. But the real carnage played out overseas. South Korea’s SK Hynix plunged over 10%, Samsung Electronics tumbled 8.73%, and a cascade of other chip-adjacent stocks cratered. Japan saw Tokyo-based Advantest lose nearly 14%, while SoftBank Group fell 7.9%. Even Europe’s semiconductor champions got hammered: Infineon Technologies shed 9.4%, ASM International dropped 9.6%, and ASML was down 6.6%.

The Financing Question No One Wanted to Ask

Here’s the uncomfortable truth that executives finally admitted: building AI infrastructure is absurdly expensive, and when borrowing costs rise, the math gets painful. Broadcom’s Hock Tan warned this week that financing could become a bigger headwind than chip shortages or power constraints. Let that sink in. We’ve spent two years obsessing over whether we could physically make enough GPUs and electricity to power AI. Turns out, paying for it all might be the real problem.

Dan Ives, tech analyst at Yorkville Ives, offered some perspective in a note that’s part reassurance, part wake-up call. He acknowledged that higher U.S. Treasury yields climbing above 5% are raising capital costs significantly. But here’s his contrarian take: this is a financing problem, not a demand problem. Hyperscalers still desperately need AI computing power. Companies still want to deploy these systems. The infrastructure buildout remains in its infancy.

“What would change our view is a demand signal, not a rate signal,” Ives said, noting that real warning signs would be actual cuts to capital expenditure plans, canceled contracts, or plummeting GPU rental prices.

Competition Gets Serious, Really Serious

Meanwhile, the competitive landscape just shifted. China’s ChangXin Memory Technologies announced a breakthrough in memory architecture that could pack more data into smaller spaces. Elon Musk chimed in on X to suggest China will be producing “a LOT of RAM in 2028.” That’s not just talk. The memory chip industry is bracing for potential oversupply as early as 2029 if manufacturers keep ramping production to meet AI demand.

This is where sentiment really sours. Investors worry that financing costs will squeeze margins on business fundamentals that already face mounting competition. The semiconductor sector thrived when supply was constrained and demand insatiable. But what happens when capacity expansion outpaces demand growth and capital becomes expensive? That’s the question keeping traders up at night.

The broader market already showed its hand Wednesday. After hotter-than-expected U.S. inflation data suggested the Federal Reserve might raise rates further, the Nasdaq Composite dropped 1.88% and the S&P 500 fell 1.11%. Tech stocks led the way down, which makes sense: they’re the most sensitive to business cycle shifts and borrowing costs.

Ives sees this as a buying opportunity for long-term believers in AI infrastructure. The pullback is tactical, not strategic. The underlying demand remains robust, he argues, and the buildout cycle is just getting started. But he’s watching for three danger signals: capex cuts, contract cancellations, or GPU price collapses.

So here we are. The AI gold rush hasn’t stopped, but the treasure hunters just realized they’re deeper in debt than they thought. The question now is whether that’s a temporary setback or a sign that we’ve been overestimating how quickly this whole thing can scale.

Source: CNBC