On January 27, 2025, the Chinese startup DeepSeek disrupted both the financial markets and the AI industry with the introduction of its open-source AI economic model. Although some analysts have expressed skepticism about the model, it poses a potential threat to the massive investments made in AI infrastructure.
DeepSeek unveiled its DeepSeek-V3 model, an intelligent assistant that has the ability to rival top market players like OpenAI’s ChatGPT. What sets the model apart is its remarkably low development cost, estimated between and million— a mere fraction of what Western companies such as OpenAI and Meta have invested. The announcement immediately impacted financial markets, with major technology giants like Nvidia and Broadcom suffering significant losses. Nvidia saw a 17% drop, losing nearly 0 billion in market capitalization, while Broadcom also dropped 17%. Other companies, including Oracle and Cisco, also faced declines of 14% and 5%, respectively.
A key feature of the DeepSeek-V3 model is that it is open-source, making it freely accessible and modifiable. This decision marks a departure from the proprietary models of many competitors and could revolutionize the AI ecosystem by enabling participants outside of China to take advantage of this breakthrough. Open systems have historically fostered innovation and lowered entry barriers, as seen with past successes like the IBM PC in the 1980s and the Android operating system in the 2010s.
DeepSeek’s approach challenges the prevailing belief in the AI industry that models only improve with larger data sets, more computing power, and increasing parameters. According to Coface, a credit insurance and business intelligence firm, if DeepSeek’s model proves effective and gains widespread industry adoption, much of the current AI infrastructure may become underutilized, leading to lower prices and a shift in investment strategies. The semiconductor industry, cloud services, and data centers could face significant challenges, as a reduction in demand for high-tech equipment may force a reevaluation of the entire value chain.
In the short term, the financial markets are expected to experience increased volatility as investors reassess the value of companies involved in AI. Major tech firms like Nvidia, Broadcom, and ASML, which supply cutting-edge semiconductor technologies, may face long-term declines in their stock value. Venture capital funds and institutional investors who have heavily invested in AI could also feel the pressure. However, while a correction is expected in the near future, the longer-term outlook could favor innovation, especially if it encourages the development of cheaper and more focused AI models.
DeepSeek’s rise comes at a time when the United States and China are engaged in an ongoing technological rivalry. Founded by Liang Wenfeng, who leads an AI center in Hangzhou, China, DeepSeek has become a symbol of China’s efforts to build a more cost-effective AI ecosystem amid increasing US restrictions on its access to advanced technology. DeepSeek’s model, which has received attention for its affordability, may also mitigate the impact of these US-imposed limitations. Although DeepSeek relies on advanced Nvidia chips, it could provide an affordable alternative for regions with high demand for AI applications but limited infrastructure, such as Europe.
Despite the excitement surrounding DeepSeek, questions remain about the hidden costs of developing such models. Expenses related to data processing, networking infrastructure, energy consumption, and the salaries of AI engineers could raise the true cost of building the model. As a result, the overall expense may be higher than initially projected.
Source: Coface
