Shares of NVIDIA, which is a major beneficiary of the AI revolution, fell 9% in the pre-market. Other tech giants, including Microsoft and Meta, lost about 4%. The Nasdaq lost 3.6%, while the S&P 500 was down 2.2%.
The decline in stock prices was not limited to the United States. European chipmaker ASML lost 9.7%, which led to a 4.8% drop in the Stoxx Europe 600 Technology index. At the same time, Japanese chipmakers Disco and Advantest fell by 1.8% and 8.6%, respectively, and China's leading chipmaker SMIC dropped by 8.4%.
Furukawa Electric, which supplies cables for data centers, lost more than 11.3%, leading the decline on the Nikkei 225 index in Tokyo.
The DeepSeek R1 model came as a surprise to the AI market, as it could be created at a much lower cost than OpenAI's models. According to NVIDIA engineer Jim Fan, DeepSeek trained its base model called V3 in two months at a cost of $5.58 million.
It is difficult to estimate the total cost of DeepSeek-R1 training; it is known that the startup used 50,000 GPUs, which could have cost hundreds of millions of dollars, so the exact numbers remain speculative. However, companies such as OpenAI, Google, and Anthropic use more than 500,000 GPUs each. As a result, investors are beginning to question whether the cost of AI chips will be as high in the future as previously predicted.
The impact of DeepSeek on NVIDIA and other hardware manufacturers could be significant. According to UBS, investments in AI by major US tech companies reached $224 billion last year and are expected to reach $280 billion in 2024. However, DeepSeek's ability to deliver similar AI performance at a fraction of the cost could put these forecasts at risk.
Despite the short-term pressures, analysts such as Dylan Patel, principal analyst at SemiAnalysis, believe that the increase in AI efficiency could have a positive impact on chipmakers in the long run. “Advancements in training and inference efficiency enable further scaling and proliferation of AI,” Patel said.