The DeepSeek Everyone Is Talking About: What Is It
On January 27, local time, U.S. artificial intelligence (AI)-related stocks were sold off due to the impact of Chinese AI startup DeepSeek, causing a historic plunge in the stock price of U.S. chip giant NVIDIA and a significant drop in the Nasdaq Composite Index.
By the close of the market, NVIDIA's stock price had fallen by 16.97%, wiping out nearly $600 billion in market value in a single day, marking the largest single-day market value loss for any company in U.S. history.
So, what is DeepSeek?
DeepSeek, officially known as Hangzhou DeepSeek Artificial Intelligence Basic Technology Research Co., Ltd., was established on July 17, 2023. It is an innovative technology company focused on developing advanced large language models (LLMs) and related technologies. In Silicon Valley, DeepSeek has long been referred to as the "mysterious force from the East" and is one of the much-discussed "Hangzhou Six Dragons" online.
Shaking Up U.S. Stocks! The Rise of DeepSeek
NVIDIA's sharp decline impacted the entire market, with Broadcom's stock price falling by 17%, Advanced Micro Devices (AMD) dropping by 6%, and Microsoft declining by 2%. Additionally, derivatives in the AI sector, such as power suppliers, were also hit hard. U.S. utility company Consolidated Edison saw its stock price fall by 21%, while Vistra's stock price dropped by 29%.
Market analysts believe that the core reason for the recent turbulence in the U.S. stock market is DeepSeek's latest breakthrough, which has shaken the dominance of the U.S. tech industry. At the end of 2024, DeepSeek released its next-generation large language model, V3, and announced it would be open-source. Test results showed that the model outperformed mainstream open-source models in multiple evaluations and had a cost advantage.
This month, DeepSeek released its latest open-source model, R1, on the opening day of the World Economic Forum's 2025 Annual Meeting, once again capturing global attention.
According to reports, the R1 model achieved significant technological breakthroughs-using pure deep learning methods to enable AI to spontaneously develop reasoning capabilities. In tasks such as mathematics, coding, and natural language reasoning, its performance rivals OpenAI's o1 model official version, with a training cost of only $5.6 million, far lower than the billions or even tens of billions of dollars invested by U.S. tech giants.
Market analysts believe that the launch of DeepSeek's model was like a "shockwave," casting doubt on the competitiveness of the U.S. tech industry and leading to the sell-off of AI-related stocks.
Jacobson, chief economist at Annex Wealth Management, stated that what truly surprised people was that companies like NVIDIA, which are considered almost monopolistic in the AI ecosystem's chips, might not have as strong a "moat" as previously thought, which could be the reason for the stock price decline.
Currently, DeepSeek's mobile application has surpassed OpenAI's ChatGPT to top the U.S. free app chart on the Apple App Store.
DeepSeek: Exceptional Performance, Excellent User Experience
DeepSeek is a large language model developed by a domestic AI company, boasting powerful natural language processing capabilities. It can understand and answer questions, assist in writing code, organize information, and solve complex mathematical problems.
Compared to OpenAI's ChatGPT, DeepSeek not only achieved performance comparable to OpenAI's o1 model but also significantly reduced the cost of inference models. Its new model, DeepSeek-R1, achieved GPT-o1-level performance at one-tenth of the cost, sparking widespread discussion in the overseas AI community.
Liu Wei, director of the Human-Computer Interaction and Cognitive Engineering Laboratory at the School of Artificial Intelligence, Beijing University of Posts and Telecommunications, explained that DeepSeek's greatest advantage lies in its algorithmic improvements and optimizations, which save on computational power and data volume. If OpenAI was previously seen as "brute force creating miracles," DeepSeek now demonstrates that "small forces can also create miracles"-small computational power using new methods can also achieve remarkable results.
Professor Yu Yang from the School of Artificial Intelligence at Nanjing University noted that DeepSeek has made corresponding optimizations in its algorithms, significantly reducing training costs.
Open Source: The Future Direction?
DeepSeek has adopted a fully open-source strategy, not only lowering the barrier to entry for users but also fostering a collaborative ecosystem within the AI developer community. Through open source, DeepSeek has attracted the attention of numerous developers and researchers, driving advancements in AI technology. NVIDIA AI scientist Jim Fan praised DeepSeek as "truly open frontier research, empowering everyone."
Of course, open source also faces challenges such as protecting intellectual property and maintaining order within the open-source community.
Professor Yu Yang stated that open source is merely a business model, and even open-source projects have copyright protections. The debate between open source and closed source essentially compares the effectiveness of different business models in specific scenarios and historical contexts.
DeepSeek's success is seen by some as a manifestation of Chinese technological idealism, even likened by foreigners to a mysterious Eastern force.
However, from the perspective of the entire AI large model industry, DeepSeek's success may represent a new development direction-improving model performance through algorithmic optimization rather than solely relying on computational power and data volume. This direction also provides new insights for the development of the AI large model industry.
