Companies & People

Nvidia shares plunge as AI infrastructure returns face scrutiny

Nvidia shares fell nearly 23% in one week amid an SEC investigation and concerns over AI business models. The selloff extended to AMD and Alphabet, while the Nasdaq Composite saw its worst weekly decline since 2020.

Nvidia shares plunge as AI infrastructure returns face scrutiny

AI infrastructure returns under scrutiny

Nvidia finished Friday at $125.53, a 47.5% drop from its October closing record of $239.24. The chipmaker lost nearly 23% during the week because reports of an SEC accounting investigation and the departure of former CEO Jensen Huang added to concerns about the AI business model. The Nasdaq Composite lost 5.3% on Friday and closed at 22,617.73, a figure 18% below its October 6 record high of 27,599.79. The 10% weekly decline for the Nasdaq is the worst since 2020, and the index has now erased its gains for 2026. The S&P 500 entered correction territory as investors questioned whether the trillions of dollars being poured into AI infrastructure will produce the expected returns.

Anthropic, the Claude developer, priced its shares at $83 during its historic $81 billion IPO, but its stock fell 9% on Friday after its first trading session closed 17% higher. CoreWeave, which rents computing capacity to AI developers, closed at $39.51, which is almost 80% below its 2025 peak. Oracle finished the week at $101.19 after it lost roughly 70% from its highs. To trigger a bubble burst on Polymarket, at least three specific events must occur within a single 90-day period, including Nvidia closing 50% below its all-time high, SOXX falling 40%, or major AI suppliers experiencing similar collapses.

Widespread selling across the semiconductor industry

AMD shed 20.5% over the week to close at $424.39, while Alphabet lost 10.3% to $265.29. The iShares Semiconductor ETF, SOXX, dropped 17.4% during the same period. This fall means SOXX has already dropped approximately 43% from its split-adjusted record closing price. Polymarket traders assess a 54% probability of a defined AI industry downturn by December 31, 2026, and a 65% probability by June 30, 2027. Investors reduced their technology exposure by the most since 2006 according to a Bank of America November fund manager survey, and 52% of those surveyed expect at least one major AI hyperscaler to announce a capital spending cut within six months.

The selloff extended into the cryptocurrency market, where Bitcoin finished the week near $58,895. Bitcoin fell 9.8% over the week, leaving the coin more than 50% below its 2025 peak. The total cryptocurrency market capitalization fell to approximately $2.01 trillion, which is a significant drop from the peak of more than $4 trillion seen previously. The Crypto Fear and Greed Index dropped to 9 and signals extreme fear. U.S. spot Bitcoin ETFs recorded $582.2 million in net withdrawals on Thursday, while forced liquidations added to the selling as Bitcoin broke below the $60,000 level.

Comparing the current crash to the dot-com era

Owen Lamont, a portfolio manager at Acadian Asset Management and former University of Chicago finance professor, told Fortune this week that he assigns a 70% probability to the U.S. stock market already being in a bubble. He tracks four warning signs: excessive valuations, widespread belief in a speculative boom, new stock issuance, and investors pouring money into the market. His assessment includes the influence of SpaceX’s massive IPO earlier this year. The European Central Bank’s November review estimates that capital expenditure by major American technology companies could reach $5.5 trillion cumulatively by 2030. This spending relies on growing debt because infrastructure spending outpaces internally generated cash.

During the original dot-com collapse, the Nasdaq lost approximately 78% between March 2000 and October 2002. The present decline is smaller, but history also offers a reminder that a technology can succeed commercially while the companies financing its expansion lose enormous amounts of market value. Richard Saperstein, chief investment officer at Treasury Partners, told Reuters the downturn is an "AI repricing" because of slowing data center construction. I find this explanation more compelling than Lawrence Gillum’s claim at LPL Financial that the decline is "not a bubble bursting" and that the AI investment cycle still has room to run. Investors are also bracing for the Federal Reserve to raise rates again in December, which increases financing costs for data centers.

Nvidia beat earnings estimates and reported record revenue and net income, yet its shares continued to collapse. The real test remains whether customers can turn computing power into profits.