Explainer: What is DeepSeek and why is it shaking up the AI industry?

The Chinese startup DeepSeek has recently made waves by launching its newest AI models, which it claims rival or even surpass industry-leading models in the United States, all at a significantly lower cost. This development threatens to disrupt the established technological hierarchy globally.

DeepSeek has garnered substantial attention in the international AI community, especially after publishing a paper last month revealing that the training of its DeepSeek-V3 model required less than $6 million worth of computational power using Nvidia H800 chips.

DeepSeek's AI Assistant, which is powered by the DeepSeek-V3 model, has outperformed its competitor ChatGPT to become the highest-rated free application on Apple's App Store in the United States. This achievement has sparked questions about the justification behind some American tech companies' decisions to invest billions of dollars in AI. The shares of several major tech firms, including Nvidia, have been adversely affected.

Here are some key details about the company that is causing a major shift in the AI sector worldwide.

WHY IS DEEPSEEK MAKING WAVES?

The release of OpenAI's ChatGPT in late 2022 triggered a rush among Chinese tech companies to develop their own AI-powered chatbots. However, when the first Chinese equivalent of ChatGPT was introduced by Baidu, a leading search engine company, it was met with widespread disappointment in China due to the noticeable gap in AI capabilities between American and Chinese firms.

The quality and cost-effectiveness of DeepSeek's models have changed this narrative. DeepSeek's two models, which have been highly praised by Silicon Valley executives and U.S. tech engineers alike, DeepSeek-V3 and DeepSeek-R1, are said to be on par with the most advanced models from OpenAI and Meta. Additionally, they are more economical to use. The DeepSeek-R1, released just last week, is reported to be 20 to 50 times cheaper to use than OpenAI's top model, depending on the specific task, according to a post on DeepSeek's official WeChat account.

However, some have publicly voiced skepticism regarding DeepSeek's success. Alexandr Wang, CEO of Scale AI, stated in a CNBC interview on Thursday that DeepSeek possesses 50,000 Nvidia H100 chips, which he claimed would violate Washington's export controls that prohibit the sale of such advanced AI chips to Chinese companies. DeepSeek did not immediately respond to requests for comments on these allegations.

On Monday, Bernstein analysts noted in a research report that the total training costs for DeepSeek's V3 model were not fully disclosed and were likely much higher than the $5.58 million the startup reported for computing power. They also pointed out that the training costs for the equally acclaimed R1 model had not been revealed.

WHO IS BEHIND DEEPSEEK?

DeepSeek is a Hangzhou-based startup controlled by Liang Wenfeng, co-founder of the quantitative hedge fund High-Flyer, according to Chinese corporate records. In March 2023, Liang's fund announced on its official WeChat account that it was "starting again," moving beyond trading to focus resources on creating a "new and independent research group to explore the essence of AGI" (Artificial General Intelligence). DeepSeek was subsequently founded later that year.

OpenAI, the makers of ChatGPT, define AGI as autonomous systems that outperform humans in most economically valuable tasks. It remains unclear how much High-Flyer has invested in DeepSeek. High-Flyer operates from an office in the same building as DeepSeek and holds patents related to chip clusters used for training AI models, according to Chinese corporate records.

In July 2022, High-Flyer's AI division announced on its official WeChat account that it owns and operates a cluster of 10,000 A100 chips.

HOW DOES BEIJING VIEW DEEPSEEK?

DeepSeek's success has already caught the attention of China's top political circles. On January 20, the day DeepSeek-R1 was publicly released, founder Liang attended a closed-door symposium for businessmen and experts hosted by Chinese Premier Li Qiang, as reported by state news agency Xinhua. Liang's presence at this event suggests that DeepSeek's achievements may be significant to Beijing's policy objectives of overcoming Washington's export controls and achieving self-sufficiency in strategic industries like AI. A similar symposium last year was attended by Baidu CEO Robin Li.