Nvidia, AMD Bounce After Global Rout—but AI's Bubble Question Won't Die

AI chip stocks clawed back some ground Friday, one day after a selloff that tore through markets from New York to Taipei. Nvidia added 2.78%, closing at $186.64, a day after it dropped 9.3%. AMD finished at $449.04, recovering from a 17.3% collapse — though both remain well shy of their October highs.

Asia took the harder hit. South Korea's KOSPI shed 6% Friday, and Taiwan's benchmark ended the week down more than 8%, its roughest stretch in months, according to Decrypt. TSMC tripped its daily limit-down threshold during the slide. Reuters reported that Taiwanese officials asked state-owned banks to step in and help support prices.

The US Tech 100 cash-market proxy bounced 1.85% to 27,007.3, still leaving it down more than 5.5% on the week. The chart shows a rebound after a steep drop from October levels — not a return to the old trend.

Politics piled on. Voters went to the polls November 3 and, in 40 of 41 local ballot contests, sided with limiting data center development, including rejecting pro-construction measures, per Ballotpedia. Some of those votes were only advisory, but the message landed. Data centers are the chip industry's biggest growth engine, and each one demands enormous spending on processors, servers, power and cooling.

The doubts didn't start at the ballot box. AI companies and their customers are spending heavily to build capacity before the returns are clear, and pricier borrowing makes that bet harder to carry. Reuters quoted strategist Steve Sosnick warning that heavily owned stocks were already “priced to perfection.”

There's a difference between a correction and a popped bubble. A correction means a drop of at least 10% from a recent peak — it says nothing about whether the underlying business is shrinking. A burst happens when expectations turn out to be fantasy and the money funding the boom dries up. The dot-com crash is the cautionary tale: the Nasdaq fell roughly 77% between March 2000 and October 2002. The internet wasn't a fad, but investors badly misjudged how fast it would pay off — telecom firms spent fortunes on fiber for customers who hadn't shown up, and network builder Global Crossing filed for bankruptcy in 2002. The technology survived. Plenty of shareholders didn't. AI could hit a similar gap between long-term promise and near-term returns, and some investors are already saying so out loud. The memory crunch has become a business opportunity in itself, with former crypto miners renting out hardware and cloud and inference startups popping up everywhere. Picture a hypothetical company borrowing billions to fill a data center with Nvidia chips, banking on renting that compute at premium rates for years. If customers spend less than expected, financing gets costlier or construction stalls — and the debt still comes due. The operator cancels chip orders, contractors lose work, utilities lose customers, and lenders eat losses. A stock selloff can stop being a market story and become an economic one. Microsoft, Alphabet, Amazon and Meta are a different animal than the dot-com startups, though — profitable, with real customers and hefty cash flows, they can absorb setbacks that would sink a smaller AI-cloud outfit borrowing to grow. When the dust settles, AI may keep remaking industries while a pile of individual AI companies fold — and investors in the winners may still have overpaid.