DeepSeek has emerged as a significant player in the AI landscape, having trained its DeepSeek-V3 model with under $6 million in Nvidia H800 chips. This achievement allowed its AI assistant to surpass ChatGPT as the top-rated free app on the U.S. Apple App Store. DeepSeek’s models are noted for their cost-effectiveness, offering substantial savings over competitors. The company’s advancements align with China’s push for self-sufficiency in AI, attracting attention from political leaders amid ongoing U.S. export restrictions.
DeepSeek’s Rise in the AI Landscape
Recently, a remarkable shift has occurred in the global AI arena, with a particular spotlight on DeepSeek. In a groundbreaking article, the company revealed that it only required an investment of under $6 million in computing resources, utilizing Nvidia H800 chips, to train its DeepSeek-V3 model. This achievement has propelled DeepSeek’s AI assistant, which is built upon DeepSeek-V3, to surpass ChatGPT, securing the title of the highest-rated free app on the Apple App Store in the United States. Such developments have sparked skepticism regarding the multi-billion dollar investments made by various American tech firms in AI, leading to fluctuating stock prices among key industry players, including Nvidia. Let’s delve deeper into what makes this innovative company a game-changer in the AI sector.
Understanding DeepSeek’s Impact
The launch of OpenAI’s ChatGPT in late 2022 initiated a flurry of activity among Chinese tech firms eager to create their own AI chatbots. The initial excitement waned, however, particularly after Baidu introduced its Chinese counterpart, leading to disappointment over the disparity in AI capabilities between American and Chinese enterprises. DeepSeek has revitalized this narrative with its highly praised models, DeepSeek-V3 and DeepSeek-R1, which are now considered competitive with the leading models from OpenAI and Meta. Notably, these models also present a more cost-effective option for users. According to information shared on DeepSeek’s official WeChat account, the recently launched DeepSeek-R1 offers a remarkable 20 to 50 times cost reduction compared to OpenAI’s model O1, depending on the specific task. Nonetheless, skepticism remains, as Scale AI’s CEO, Alexandr Wang, raised concerns about DeepSeek’s claimed resources, alleging they possess 50,000 Nvidia H100 chips while suggesting potential issues with compliance to U.S. export regulations. DeepSeek has yet to respond to these claims, and analysts from Bernstein have indicated that the actual training costs for the V3 model may be significantly higher than reported.
The Visionaries Behind DeepSeek
DeepSeek, a startup located in Hangzhou, is primarily owned by Liang Wenfeng, a co-founder of the quantitative hedge fund High-Flyer. In March 2023, Liang’s fund announced a strategic shift towards establishing a dedicated research group aimed at exploring the foundations of Artificial General Intelligence (AGI). DeepSeek emerged shortly after this announcement. AGI, as defined by OpenAI, refers to systems that can outperform humans in most economically significant tasks. While the exact level of investment from High-Flyer into DeepSeek is unclear, records indicate that the two companies share office space and that High-Flyer holds patents for chip clusters utilized in training AI models. In July 2022, High-Flyer’s AI division revealed that it managed a cluster of 10,000 A100 chips.
DeepSeek’s Political Significance in China
The success of DeepSeek has not gone unnoticed within influential political circles in China. On January 20, the same day DeepSeek-R1 was unveiled, founder Liang participated in a private symposium attended by Chinese Premier Li Qiang, signaling the importance of DeepSeek’s innovations to Beijing’s ambitions. This involvement might suggest that the company’s advancements are aligned with China’s objectives to become self-sufficient in strategic sectors like AI, particularly in light of ongoing challenges posed by U.S. export controls. Notably, Robin Li, CEO of Baidu, was present at a similar symposium last year, indicating a growing recognition of AI’s pivotal role in China’s future.