Latest Cover Story | The Sudden Rise of DeepSeek Shakes the Fundamental Logic of AI Investment and Computing Power Competition (AI Translation)

最新封面报道｜DeepSeek爆火 撼动AI投资和算力竞争底层逻辑

A seismic shift in the U.S. capital markets was triggered by the rise of DeepSeek, a Chinese artificial intelligence company under Quantum Hedge Fund, headquartered in Hangzhou, China. Following the unveiling of its next-generation open-source model DeepSeek-V3 in December 2024, the U.S. stock market saw a major upheaval as financial institutions swiftly sold shares in chip companies, causing a significant decline in Nvidia’s market valuation[para. 1]. DeepSeek’s free and highly accessible AI app has quickly risen to the top of the U.S. App Store, outpacing well-established apps like ChatGPT, Temu, and Shein, sparking discussions on social media[para. 3].

The model’s unprecedented rise and open-source innovation have increased traffic to DeepSeek's website by 2257% month-on-month, hitting 278 million visits in January 2025, a noteworthy achievement even if it still trails ChatGPT’s 3.8 billion visits[para. 4]. DeepSeek’s breakthrough comes at a time when OpenAI’s focus has shifted from pre-training to enhancing model capabilities with reinforcement learning, motivating the introduction of its o1 model[para. 6]. Experts note that DeepSeek-V3’s capabilities are comparable to those of GPT-4o, pointing out that while DeepSeek’s training costs are significantly lower, its performance is on par with leading models like those from OpenAI and Meta’s advanced models[para. 9].

Industry leaders such as Microsoft CEO Satya Nadella and Scale AI’s Alexandr Wang acknowledge DeepSeek's potential, noting its cost-efficient approach to achieving high-performance AI[para. 12][para. 13]. DeepSeek’s low computational cost has provoked skepticism about whether massive investments in computational power truly offer a lasting competitive advantage, as the emergence of the Chinese company has proved otherwise[para. 15]. On the trading floor, Nvidia and several major U.S. chip companies experienced significant value losses, instigated by DeepSeek’s breakthrough, marking the largest one-day market decline for Nvidia in history[para. 16][para. 17].

Despite allegations of data misuse and appropriating OpenAI’s models, such claims lack evidence and face dismissal from AI experts who see platform models as being trained on publicly available web data[para. 19][para. 23]. DeepSeek has quickly become a focal point in the U.S.-China AI competition—past U.S. presidential policies imposed semiconductor export controls impacting China’s access to AI chips, a narrative DeepSeek’s open-source model stands to disrupt[para. 25][para. 27].

DeepSeek's rise highlights the importance of efficient AI development amidst an industry largely dominated by proprietary models. The company’s technique of scaling reinforcement learning shows a potential paradigm shift away from closed-source approaches. Industry professionals like Tsinghua University's Liu Zhiyuan note the strategic necessity for Chinese firms to embrace open source and algorithm optimization to remain competitive, advocating for the creation of innovative mechanisms and a better software market environment within China[para. 34][para. 36].

Overall, DeepSeek has triggered a reevaluation of global AI strategies, underscoring the dynamism and unpredictability inherent in technological advancement[para. 38]. Its emergence signals not only a potential title shift in AI leadership but also highlights the crucial role cooperation and open-source strategies may play in future AI development[para. 40].