The Nasdaq Composite fell by almost three percent in the past trading week, while the Philadelphia Semiconductor Index even dropped by 8.5 percent. Investors are increasingly doubting whether the billion-dollar investments of major tech companies in AI infrastructure will yield the expected returns. Chip manufacturers like Micron and Kioxia lost between a quarter and half of their market value within a few weeks.
Semiconductor stocks lose more than the broader market
The sell-off hit chip and memory manufacturers particularly hard. The Philadelphia Semiconductor Index is now about 19 percent below its June high after the weekly loss of 8.5 percent. As reported by the Irish Times, this is the strongest weekly decline since the sell-off caused by tariffs in April 2025. The memory chip manufacturer Micron alone lost about 25 percent of its market value in July. Sandisk and Western Digital each lost more than 30 percent during the month. The Japanese Kioxia fell by 16 percent in one week and is more than 50 percent down from its June high. Contract manufacturers TSMC and ASML also declined by seven and 4.6 percent, respectively. In Asia, the Japanese Nikkei 225 fell by four percent, and the Chinese CSI 300 by 3.6 percent. The Chinese AI providers Z.ai and MiniMax lost 28.5 and 15.6 percent on the stock market, respectively. The decline thus affected not only U.S. stocks but the entire global supply chain around AI data centers – from memory chips to contract manufacturers.
Analysts warn of debt and lack of demand
Behind the price drop lies a fundamental question: Are the billion-dollar expenditures of hyperscalers for building AI data centers already paying off today? Among the largest investors are Alphabet, Amazon, Meta, Microsoft, and Oracle – whether the expenditures will pay off is likely to become clear only in a few years. According to CBS News, economist Kate Brennan from the AI Now Institute observes that companies are increasingly financing themselves through debt markets because the expected returns have so far not materialized. Market strategist Ed Yardeni from Yardeni Research puts it succinctly: The AI ecosystem is at risk of wobbling if the hoped-for end-user demand does not materialize. Vanguard economists Qian Wang and Kevin Khang compare the current situation to the dot-com bubble of the early 2000s and expect a bumpy phase in the markets. Apollo analyst Torsten Sløk warns that disappointing revenues from hyperscalers in the cloud and AI business could burden the entire stock market. If companies significantly reduce their data center investments as a result, it could even dampen the entire U.S. economy.
Goldman Sachs estimates investments at $7.6 trillion
How high the stakes actually are is shown by an analysis from Goldman Sachs from May of this year. The investment bank estimates global spending on computing power, data centers, and energy supply between 2026 and 2031 to total around $7.6 trillion. Annual investments are expected to rise from about $765 billion this year to $1.6 trillion in 2031, more than doubling the spending level within five years. The Goldman authors George Lee and Lucas Greenbaum also point out the uncertainty of the forecast. It heavily depends on assumptions regarding the lifespan of chips, the construction costs of data centers per megawatt, and the pace of potential supply bottlenecks. The figure is therefore significantly more uncertain than it appears at first glance, the authors themselves write. The forecast cannot be independently verified, as it is based on the bank’s internal model assumptions. Nevertheless, the study provides investors with a benchmark against which future investment announcements from hyperscalers can be measured.
Not all analysts see fundamental causes
Not every market observer interprets the sell-off as a loss of confidence in AI business models. HSBC strategist Max Kettner attributes part of the movement to the abrupt end of leveraged bets on continued price gains, describing it as a “brutal unwinding” of speculative positions. Fund manager Hao Hong from Lotus Asset Management speaks of a technical sell-off by quantitative funds and not a fundamental reassessment of the industry. Michael Zigmont from the Visdom Investment Group notes that some investors are simply looking for a reason to sell in light of solid quarterly numbers. UBS Wealth Management remains more optimistic. Its strategist Charlie Anderson expects the S&P 500 to rise to 7900 points by the end of the year despite the current turbulence.
It will be crucial whether the hyperscalers can present solid revenue figures from their AI services in their upcoming quarterly reports. If concrete revenues are lacking, the debate over a possible overinvestment in AI infrastructure is likely to intensify – especially since an increasing portion of the expansion is being financed through debt.
