DeepSeek: Revolutionizing AI Competitiveness and China's Strategic Ambitions

The Emergence of DeepSeek: A Game-Changer in AI Innovation

BEIJING – The AI sector is observing a significant shift with the recent introduction of Chinese startup DeepSeek's cutting-edge artificial intelligence models. Claimed to rival or surpass leading U.S. models at a more economical cost, DeepSeek is poised to redefine the technology landscape.

With its global reputation growing, DeepSeek gained attention through a paper that revealed its DeepSeek-V3 model required under US$6 million in Nvidia H800 computing power for training. This breakthrough has propelled DeepSeek's AI Assistant to overtake ChatGPT as the top-rated free app on the U.S. Apple App Store, sparking questions regarding the large-scale investments by U.S. tech companies and impacting major tech shares, including those of Nvidia.

Why Is DeepSeek Attracting Global Attention?

The advent of OpenAI's ChatGPT in late 2022 spurred a wave of AI development among Chinese technology firms. However, the initial disappointment over the performance gap between Chinese and U.S. AI technologies, led by Baidu, has been overturned by DeepSeek's cost-effective and high-quality models.

The DeepSeek-V3 and DeepSeek-R1 models are comparable to the advanced models from OpenAI and Meta, with V3 being particularly lauded for its affordability. As detailed on DeepSeek's WeChat account, DeepSeek-R1 is reported to be 20 to 50 times more cost-effective than OpenAI's leading models.

Despite this acclaim, some skepticism remains. Scale AI's CEO, Alexandr Wang, without evidence, claimed DeepSeek's possession of 50,000 Nvidia H100 chips could breach U.S. export restrictions, but DeepSeek has not commented on the claim. Additionally, Bernstein analysts argued that the actual training costs for DeepSeek's models could exceed the reported US$5.58 million, also noting that costs for the R1 model remain undisclosed.

The Visionary Minds Behind DeepSeek

Founded in Hangzhou, DeepSeek's major stakeholder, Liang Wenfeng, is renowned as the co-founder of High-Flyer, a quantitative hedge fund. High-Flyer, which announced a strategic pivot in March 2023, is focusing on developing Artificial General Intelligence (AGI), defining it as autonomous systems outperforming humans in numerous economically valuable tasks, as per OpenAI's definition.

The extent of High-Flyer's financial commitment to DeepSeek is not fully publicized, though they share an office and patents for AI model training chip clusters, a key component in DeepSeek's development arsenal.

DeepSeek's Significance in China's Strategic Agenda

The remarkable rise of DeepSeek has captured the attention of China's political elite. Founder Liang Wenfeng's presence at a private symposium with Chinese Premier Li Qiang on January 20, the day DeepSeek-R1 released, signals its potential impact on Beijing's strategic aims. China's stance on overcoming U.S. export controls and achieving technological self-reliance is underscored by DeepSeek’s achievements, paralleling last year’s event featuring Baidu CEO Robin Li.