DeepSeek and China’s AI Ambitions

Reader’s note: Happy Lunar New Year! As we welcome the Year of the Snake, we here at Sino Strategies are embracing the future with cautious optimism and unbridled ambition. The new year brings a fresh start and many important developments in China’s global influence.

To our incredible readers, thank you for being part of our journey. Your support and engagement fuels our work and keeps us motivated to offer strategic insights that help navigate the ever-evolving U.S.-China relationship. So from the bottoms of our hearts, 谢谢你!

Summary

  • Founded in 2023, DeepSeek gained prominence by driving innovations in AI model efficiency with older GPUs. Its success highlights China’s AI advancement and has unsettled U.S. markets.
  • DeepSeek’s growth raises concerns over tech competition, national security, and export control effectiveness, and is likely to prompt stricter regulations. U.S. businesses may need to enhance domestic AI innovation capabilities and workforce development to remain competitive.

DeepSeek, the High-flying Chinese AI company

DeepSeek was founded in 2023 by hedge fund manager Liang Wenfeng. A self-described AI-enthusiast, he had established himself in the world of finance as an investor who developed and deployed AI trading algorithms for profit. 

DeepSeek’s main innovation centers around reduced memory usage and computational power in training its AI models. Indeed, the company claims to have used Nvidia GPU chips that are over two years old (i.e. a lifetime in tech industries) to train the model. Indeed, starting in 2021, and over the course of a period pre-dating the U.S.’s imposition of export controls on American semiconductors, founder Liang Wenfeng stockpiled GPU chips for “an AI project.” Indeed, DeepSeek has mostly been set up as a research company, rather than a corporation meant to compete with large incumbent players in the U.S. or, for that matter, China. Nevertheless, DeepSeek has taken the world — and especially the U.S. stock market — by storm. 

China’s Global AI Supremacy: Fact or Fiction?

While DeepSeek’s sudden and stunning rise may have rattled the U.S. stock market, the company’s rise is not that unexpected when considered from the perspective of China’s AI development ecosystem and overall tech industry landscape. The CCP has invested billions of dollars into and supported the development of a flourishing tech R&D environment around AI over the past several decades. Indeed, one of the Chinese Ministry of Science and Technology’s “national champions” — Alibaba — recently launched an updated version of its Qwen-2.5 model to rival U.S. competitors and DeepSeek alike. 

DeepSeek’s meteoric rise has not been without question, however. Despite the company’s apparent compliance with semiconductor export controls and utilization restrictions, David Sacks — Chair of the Council of Advisors on Science and Technology under President Trump — claimed there is “substantial evidence” that DeepSeek distilled knowledge out of one of OpenAI’s ChatGPT models. Furthermore, from a U.S. national security perspective, DeepSeek’s widespread penetration throughout the U.S. may pose data and privacy infringement concerns, similar to the concerns posed by TikTok. As such, the U.S. may seek to restrict China’s global AI influence even more aggressively in the wake of DeepSeek’s success, further limiting the CCP’s global ambitions for AI supremacy. 

Implications for the U.S.

  1. U.S. Policy Implications

On the one hand, DeepSeek’s success reveals some of the challenges related to successfully implementing export controls to curb technology development abroad. Arguably, the United States’s semiconductor export controls may have facilitated DeepSeek’s rise, forcing the tech company to innovate with the resources that were readily accessible. Other companies such as Huawei and SMIC may have had similar experiences. 

On the other hand, export controls are not an overnight solution, and it is still too early to tell if the controls are limiting China’s ability to access compute power. Indeed,  the controls are less about slowing or preventing individual projects than about constraining China’s ability to build and maintain a robust, scalable, and self-sustaining cutting-edge AI ecosystem. From this perspective, and given the Trump Administration’s hardline stance on international trade and foreign policy, it seems likely that more stringent hardware export restrictions are on the horizon. The Trump administration’s response to DeepSeek’s rise could likely also involve controlling open source software, a challenge for which there is no easy solution. 

  1. U.S. Business Implications

For U.S. businesses, the successful rise of a Chinese AI company is clearly a sign of increased international competition. U.S. players will likely continue to face increasingly steep competition from Chinese companies, and others around the world, in AI. To remain competitive, U.S. technology companies could start by reviewing DeepSeek’s open source models and identifying potential algorithmic deficiencies in their own software. 

Additionally, U.S. companies could seek to align more closely with the U.S. government on investments related to growing the American AI and technology workforce through increased educational opportunities and improved talent acquisition, retention, and development programs. One potential drawback of heightened focus on export controls and national security is the cultural blowback such a focus could have, preventing the U.S. from attracting top talent globally and collaborating openly with other countries on AI initiatives. U.S. AI companies may be tasked with maintaining global collaborative spirit on AI projects. 

  1. U.S. Consumer Implications

For the last several years, growth drivers of the U.S. public securities industry have been American tech and AI stocks. DeepSeek’s success, and the associated dropoff in the U.S. stock market, shows that there are systematic risks in investment strategies that rely solely — or even heavily on — said stocks. Nevertheless, markets always tend to react emotionally to global developments, and DeepSeek’s competition — including many U.S. tech companies — are going to respond competitively to its success, including improving the training efficiency of their own AI models. Long term, AI tech is likely still a growth opportunity in the U.S. — and, as DeepSeek has shown, around the globe. 

2 responses to “DeepSeek and China’s AI Ambitions”

  1. […] The white paper emphasizes security in three main areas: political security, territorial & international security, and development & innovation security. These have all been central themes of Xi Jinping’s leadership, demonstrated by the steps he has taken to purge corruption throughout all areas of the CCP, spread China’s influence regionally and internationally, and promote domestic innovation and development in key emerging fields and critical technologies.  […]

    Like

  2. […] an AI chatbot rivaling the highest performing U.S.-based generative AI models at the time for, presumably, only a fraction of the training and computing resources. However, DeepSeek is not considered one […]

    Like

Leave a comment