DeepSeek: An Unlikely Game Changer in the US-China AI Conflict

The DeepSeek logo on a phone screen / Reuters-Yonhap

DeepSeek, a Chinese start-up, has made a dramatic impact on the global tech landscape just before the Lunar New Year. This company has unexpectedly caused the stock prices of major players in the AI industry, like semiconductor giant Nvidia, to plummet. Their recent breakthroughs in AI have introduced new challenges to established tech companies.

Based in Hangzhou, DeepSeek has introduced two advanced AI models, DeepSeek-V3 and DeepSeek-R1, in just a few weeks. Their innovative approach allowed them to create these models with significantly less financial and computational resources than those typically used by tech giants to develop large language models (LLMs), which are essential for generative AI applications such as ChatGPT.

Many in the AI community have praised DeepSeek’s open-source models, noting their impressive performance in industry benchmarks. This is particularly remarkable given the recent U.S. restrictions limiting China's access to cutting-edge semiconductor technology. Sam Altman, CEO of OpenAI, shared his admiration for DeepSeek's capabilities on social media, applauding the value they deliver.

The excitement around DeepSeek peaked when their chatbot, which incorporates the R1 reasoning model, rose to the top of the free apps category on Apple's App Store in both the U.S. and China.

Now, let's explore what factors contributed to DeepSeek's success:

A Visionary at the Helm

Liang Wenfeng, the founder and CEO of DeepSeek, is not a public figure, but he has shared his insights through interviews with tech media in China during 2023 and 2024. When he launched DeepSeek in May 2023, he envisioned pursuing artificial general intelligence (AGI) and committed to making the outcomes of this research accessible to everyone. Liang emphasized the importance of democratizing AI technology rather than allowing it to remain under the control of a select few companies.

Liang’s goal with DeepSeek was not just to contribute to existing technologies but to expand the boundaries of what AI can achieve. He expressed frustration that Chinese companies were not more actively involved in innovative research, stating that China should contribute to global developments rather than merely benefiting from them.

A Talented Team

DeepSeek's team consists mainly of young Chinese scientists, many of whom graduated from top universities in China, including Tsinghua University and Peking University. The team includes both fresh graduates and PhD candidates, showcasing a blend of innovative ideas and youthful energy.

The technical paper for the R1 model has about 200 co-authors, highlighting the collaborative effort that has gone into its development. One notable former DeepSeek employee, Luo Fuli, has recently garnered attention for her talent, reportedly being offered a 10 million yuan compensation package from Xiaomi, though she has not accepted yet.

Funding and Resources

Though DeepSeek has a modest registered capital of just 10 million yuan, it benefits from the financial backing and technological resources of High-Flyer Quant, a leading quantitative hedge fund founded by Liang himself.

Despite challenges posed by U.S. export restrictions on advanced chips, Liang has managed to accumulate a significant number of graphics processing units from Nvidia. High-Flyer Quant's focus is on exploring AGI and has created a separate entity, DeepSeek, to focus on this pioneering research with a commitment to substantial investment.

An Agile Management Approach

DeepSeek operates with a streamlined structure that allows for rapid decision-making and resource allocation. Liang noted that there are no bureaucratic barriers within the company, which aids in quick responses to technological developments.

Liang's management emphasizes a natural division of labor, allowing each employee to contribute according to their strengths and ideas, fostering a creative and innovative work environment.

The story of DeepSeek epitomizes how emerging start-ups can disrupt traditional tech giants through innovation and agile strategies, showcasing a new era of AI development.

DeepSeek, AI, Innovation