The launch of Chinese AI startup DeepSeek’s latest AI model, which has been hailed as outperforming its U.S. counterparts, caused a significant stir in global markets, leading to a sharp decline in tech stocks across the U.S. and Europe.
The new model has been praised for its cost-efficiency and superior performance compared to top U.S. models, raising doubts about the overvaluation of tech stocks.
Shares of Nvidia, the world’s largest AI chipmaker, plummeted by 17%, erasing $600 billion (€573 billion) in market capitalization—marking the largest intraday decline in its history. European AI chip equipment manufacturer ASML saw its stock drop by 7%. The U.S. tech-heavy Nasdaq Composite shed over 3%, and the tech sector as a whole saw a 5.6% drop. European stock markets followed suit, weighed down by losses in the tech sector.
The market turbulence also affected the U.S. dollar, which weakened against other G-10 currencies as U.S. government bond yields sharply declined. Meanwhile, the Euro surged to a six-week high against the dollar. Commodities like gold and crude oil saw significant drops as investors sought safer assets amidst the market turmoil. However, the dollar began to rebound in Tuesday’s Asian session, causing pullbacks in other currencies.
Kyle Rodd, senior market analyst at Compital.com, suggested that the market reaction reflected a “shoot first and ask questions later” approach. “There’s a lot of information to parse and discount,” he noted, particularly in regards to the credibility of DeepSeek’s performance claims.
DeepSeek, founded in 2023 by the hedge fund High Flyer, unveiled its free, open-source large language model (LLM) R1 in late December. The model was developed in just two months at a cost of under $6 million (€5.7 million), far less than the billions invested by U.S. hyperscalers in AI infrastructure. The Chinese LLM has been particularly praised for outperforming high-profile systems like OpenAI’s GPT-4, Google’s Gemini 2.0 Flash, Anthropic’s Claude 3.5 Sonnet, and Meta’s Llama 3.1, especially in solving complex mathematical and coding problems.
On the day of its release, DeepSeek’s AI Assistance app became the most downloaded free app on the Apple App Store, surpassing ChatGPT. However, the company temporarily halted registrations later in the day after experiencing a large-scale cyberattack on its services.
DeepSeek’s model was trained using Nvidia’s less advanced H800 AI chips, despite the U.S. tightening AI chip exports to China, limiting access to higher-end products. In response, Nvidia acknowledged DeepSeek’s achievements but emphasized that running AI models still requires a substantial amount of its products. The company also confirmed that DeepSeek complies with U.S. export control laws.
