Cover Story: How DeepSeek Is Redefining the Future of AI

DeepSeek, a Chinese AI startup, has quickly risen to prominence by achieving AI capabilities comparable to top models like those of OpenAI, but with much greater efficiency. This development has resulted in upheaval within the global tech industry, sparking sell-offs in chip stocks and raising political tensions in both the U.S. and abroad. DeepSeek's consumer app has surged to the top of app stores, forcing U.S. policymakers and investors to reconsider their stance on AI technology [para. 1].

The startup, supported by High-Flyer, a private equity firm in Hangzhou, introduced its model DeepSeek-V3 at the end of 2023, followed by its reasoning model R1. These models, which are highly resource-efficient, rival OpenAI's existing models, leading to widespread public interest in DeepSeek's free platform as opposed to OpenAI’s subscription-based service [para. 2][para. 3].

DeepSeek's models require far less computational power, using only 2,048 Nvidia GPUs compared to the extensive resources OpenAI employs. The entire training cost for DeepSeek was significantly lower, leading to market shifts as chip manufacturers saw devaluation due to the presumed reduced demand for high-end chips [para. 4][para. 5].

Microsoft's CEO highlighted the importance of recognizing China's advancements in AI during the World Economic Forum in Davos, positioning DeepSeek's progress as an example of engineering ingenuity on par with or exceeding that of American firms [para. 6]. However, it has also led to calls within the U.S. for stricter controls on AI exports to maintain technological dominance [para. 7].

Accusations have surfaced suggesting that DeepSeek might have used OpenAI's models illicitly, but these claims lack robust evidence. Many experts argue that AI models largely rely on publicly available datasets for development, complicating claims of intellectual property infringement [para. 8].

High-Flyer, the parent company of DeepSeek, has dedicated extensive resources toward AI since 2015. While other companies have struggled to secure computing resources, High-Flyer had anticipated and stockpiled these resources earlier. DeepSeek's focus on open-source models allowed for groundbreaking developments in AI efficiency and cost reduction, compelling other tech giants like ByteDance and Alibaba to adjust their strategies [para. 9][para. 10][para. 11].

The rise of DeepSeek has created a ripple effect internationally, prompting cloud service providers and chip manufacturers to quickly adapt to DeepSeek's models. Companies like Microsoft and Amazon have integrated DeepSeek models into their services, and local Chinese cloud providers are utilizing and enhancing DeepSeek’s open-source technologies to remain competitive [para. 13][para. 14][para. 15][para. 18].

Debates in the AI community have intensified over the sustainability of large-scale AI infrastructure investments. Some suggest spending on AI might be bloated if models like DeepSeek's can achieve high efficiency with less resources. However, big tech companies argue that inference models still require significant resources, thus maintaining their substantial investments despite potential efficiency gains in the model training process [para. 37][para. 39][para. 40].

DeepSeek faced a significant cyberattack in January 2025, believed to be a large-scale DDoS campaign, highlighting challenges in the geopolitical tech space. Allegations of misusing OpenAI data and illicitly acquiring U.S. chips further spotlighted the frictions between China and the U.S. regarding AI advancement [para. 48][para. 49][para. 50].

Despite these challenges, DeepSeek continues its trajectory by leveraging its open-source model, aiming to further innovation and collaboration in the AI sector, rather than relying solely on proprietary advances—a strategy that could disrupt existing business models of firms like OpenAI and Anthropic [para. 56][para. 63].