Chinese AI startup DeepSeek has recently unveiled its latest artificial intelligence model, DeepSeek-V3, marking a significant advancement in the AI landscape. This new model employs a mixture-of-experts architecture, enabling it to activate only the necessary parameters for specific tasks, thereby enhancing efficiency and performance. With 671 billion parameters, DeepSeek-V3 outperforms leading open-source models, including Meta’s Llama 3.1-405B, and closely matches the performance of proprietary models from companies like Anthropic and OpenAI.
One of the most notable aspects of DeepSeek-V3 is its cost-effective training process. The model was trained using approximately 2,000 Nvidia H800 GPUs over 55 days, at a cost of around $5.58 million. This expenditure is roughly one-tenth of what U.S. tech giant Meta invested in developing its latest AI technology.
The release of DeepSeek-V3 has had a profound impact on the global tech industry. Following its launch, major technology companies experienced significant stock declines. Nvidia’s stock fell by as much as 17–18%, while other tech firms, including Microsoft, Alphabet (Google’s parent company), and Dutch chip equipment maker ASML, also saw substantial decreases. This market reaction led to record losses of about $593 billion in the market capitalizations of AI and computer hardware companies, with a total of $1 trillion wiped off American stocks by January 28, 2025.
Industry leaders have expressed mixed reactions to DeepSeek’s advancements. Microsoft CEO Satya Nadella and OpenAI CEO Sam Altman have both praised the model as “super impressive.” In contrast, figures like Elon Musk have expressed skepticism regarding the app’s performance and the sustainability of its success.
DeepSeek’s founder, Liang Wenfeng, has been compared to OpenAI CEO Sam Altman, with CNN referring to him as the “Sam Altman of China” and an evangelist for AI. Chinese state media have widely praised DeepSeek as a national asset. On January 20, 2025, China’s Premier Li Qiang invited Wenfeng to a symposium with experts, seeking his opinions and suggestions on a draft of the annual 2024 government work report.
DeepSeek’s optimization of limited resources has highlighted potential limits of United States sanctions on China’s AI development, which include export restrictions on advanced AI chips to China. The success of the company’s AI models has consequently “sparked market turmoil” and caused shares in major global technology companies to plunge.
In summary, DeepSeek’s launch of the DeepSeek-V3 model represents a significant milestone in AI development, demonstrating that high-performance AI can be achieved with efficient resource utilization. This development has not only disrupted global markets but also prompted a reevaluation of strategies within the AI industry.
