The stock prices of leading tech companies came under pressure on Monday after the Chinese AI start-up DeepSeek raised doubts about the sustainability of billion-dollar investments in AI hardware in the US with the release of its latest large language model.
DeepSeek claims to train its models with significantly fewer Nvidia chips than U.S. competitors, raising questions about the future need for extensive hardware purchases. The chatbot developed by DeepSeek, a direct competitor to OpenAI's ChatGPT, reached number one on the download charts in Apple's U.S. App Store over the weekend.
The news caused Nvidia's stock to fall by 14 percent in pre-market trading, while Microsoft and Meta declined by 7 and 5 percent, respectively. Nasdaq futures indicated a drop of 4.4 percent, while the S&P 500 fell by 2.5 percent. European tech stocks also came under pressure: ASML's stock, a leading semiconductor equipment manufacturer, fell by more than 10 percent, and the STOXX Europe 600 Technology Index lost 5.2 percent.
It is clearly DeepSeek," said a fund manager in Tokyo, referring to the uncertainty over whether investments in AI hardware would actually pay off to the expected extent. According to UBS, investments by major US tech companies in AI infrastructure amounted to $224 billion last year and are expected to rise to $280 billion by 2024. At the same time, OpenAI and SoftBank had recently announced plans to invest up to $500 billion in AI infrastructure over the next four years.
The AI trade is vulnerable if it is based on the assumption that the technological lead remains inviolable," commented Luca Paolini, chief strategist at Pictet Asset Management.
DeepSeek, founded by hedge fund manager Liang Wenfeng, made waves last week with the release of a detailed report on its cost-efficient training methodology. U.S. venture capitalist Marc Andreessen compared it on platform X to the "Sputnik moment" of artificial intelligence – alluding to the technological wake-up call for the U.S. by the Soviet Union in the 1950s.
However, analysts warned of an overreaction by the market. Dylan Patel, chief analyst at SemiAnalysis, emphasized that efficiency improvements could even have positive long-term effects for hardware manufacturers like Nvidia. "Efficiency gains in training and inference ultimately promote the scaling and dissemination of AI," said Patel.
While Western markets recorded losses, Chinese tech stocks in Hong Kong rose: Baidu gained 4 percent, Alibaba increased by 3 percent. However, the Chinese CSI 300 Index fell slightly by 0.4 percent.
