DeepSeek: The Chinese Chatbot That Wants to Compete with ChatGPT

Chinese AI startup DeepSeek launched its AI Assistant powered by the DeepSeek-V3 model, which quickly overtook OpenAI's ChatGPT to become the top-rated free application on the Apple App Store in the United States. This achievement, alongside claims that DeepSeek's models outperform or match US AI capabilities at a fraction of the cost, has significantly shaken the AI sector.

“The DeepSeek-V3 model required less than US$6 million worth of computing power from Nvidia H800 chips,” says DeepSeek. This claim challenges the prevailing market belief that AI growth would continue to drive demand for high-tech chips, primarily from companies like Nvidia, and cast doubt on the sustainability of US dominance in the AI space, reports Reuters.

DeepSeek’s rise comes amid the ongoing global AI race. The launch of OpenAI’s ChatGPT in late 2022 set off a wave of competition, particularly in China, where local companies sought to develop their own AI chatbots. While early models failed to meet expectations, DeepSeek’s DeepSeek-V3 and DeepSeek-R1 models have exceeded performance benchmarks set by US firms like OpenAI and Meta.

The DeepSeek-R1 is reportedly 20 to 50 times less expensive to operate than OpenAI’s models, depending on the task, while maintaining a performance on par with OpenAI-01. According to DeepSeek’s claims, the total training cost of the DeepSeek-V3 model was less than US$6 million, while models developed by major US firms typically cost over US$100 million to train.

Some, however, are skeptical of DeepSeek’s capabilities. Alexandr Wang, CEO, Scale AI, has questioned the scale of DeepSeek’s operations, claiming that the startup might be using as many as 50,000 Nvidia H100 chips, a quantity that could potentially violate Washington’s export controls. These concerns have been compounded by the lack of transparency regarding DeepSeek’s actual training costs and chip usage. Despite these uncertainties, the growing popularity of DeepSeek’s models in the US suggests that its cost-effective approach is gaining traction.

The emergence of DeepSeek poses several implications for the broader AI sector. First, it highlights the potential for more cost-efficient AI models, which could make advanced AI applications more accessible to a wider range of companies, especially those in industries with tight budget constraints. Lower training costs, according to Reuters, could also foster faster innovation, as more players in the market would be able to access the tools needed to develop new AI solutions.

Second, the success of DeepSeek raises questions about the effectiveness of current tech export controls aimed at limiting China’s access to advanced AI chips. While these controls were designed to curb China’s ability to develop cutting-edge AI models, the affordability of DeepSeek’s models suggests that China may have found a way to work around these restrictions. 

Finally, DeepSeek’s rise is forcing US tech companies to reconsider their investment strategies. Companies like Nvidia, Microsoft, Meta, and Alphabet have seen their stock prices drop, as investors question the viability of betting on the high demand for chips and data infrastructure when cheaper alternatives are emerging. The selloff in tech stocks, particularly those associated with AI infrastructure, signals a shift in investor sentiment, with hedge funds becoming more cautious about continuing to pour capital into US AI-related companies. According to Goldman Sachs, hedge funds have been reducing their bets on US tech stocks, with a particular focus on companies in the AI infrastructure sector.

However, DeepSeek’s sudden success has been accompanied by technical challenges, including a cyberattack that temporarily impacted its registration process. Despite these setbacks, the company’s AI assistant quickly became the top-rated free app on the Apple App Store in the United States.