DeepSeek “Sputnik” Breakthrough: Large Language Models in a Techno-Geopolitical Contestation between the U.S. and China

  • 05/03/2025
  • Category: Commentaries
No Comments
deepseek

Author: Emira Anjani
Editor: Hafiz Noer

In just a matter of weeks, the DeepSeek R1 model has taken the world by storm, disrupting the dominance of ChatGPT-4o as the leading Large Language Model (LLM). Its recent launch has propelled the intensification of the U.S.-China technological competition, reshaping the landscape of AI innovation. DeepSeek was founded by Liang Wenfeng in 2023 under the operation and management of a Hangzhou-based technology company.[1] In terms of its functionality, DeepSeek is primarily viewed as the equivalent of ChatGPT–though some have argued that the former is particularly ideal for technical problem-solving and domain-specific tasks, while the latter is more suitable for day-to-day use and creative tasks. Users have also said that DeepSeek’s natural language output is more contextual and task-focused, whereas ChatGPT is friendly and conversational; this factor might account for users’ preferences, though not so much to determine comparative advantage.[2] Regardless of which model holds the operational advantage, the two-year-long heyday of ChatGPT as the leading global LLMs is seemingly facing, or is about to, an end with the counterintelligence of DeepSeek being the alternative for an advanced AI chatbot.

From the standpoint of geopolitics and technological competition, the emergence of DeepSeek embodies a shift in the global AI landscape from unipolarity to bipolarity. Above all, however, this shift indicates China’s success in maintaining a power balance against the U.S. in their ever-intensifying scramble for global leadership. To a certain extent, this shift resembles another ‘Sputnik’ moment for the U.S. that disrupts and threatens its techno-geopolitical standing. Given these circumstances, this article seeks to examine further how DeepSeek challenges U.S.’s AI dominance. The analysis unfolds in two parts: first, it provides a conceptual overview of techno-geopolitics, particularly that of the rivalry between the U.S and China. This section highlights that AI has become an object of competition in the complex relations between the two great international powers. Then, it is followed by a closer look into the ways in which China’s DeepSeek disturbs the U.S.’s AI dominance, focusing on two key aspects: the American business model and the democratization of AI capabilities.

AI as a Pawn of Techno-Geopolitics

Wong (2021) argues that beneath the U.S.-China economic rivalry is a “more deep-seated competition for structural power” for a global social and political influence, with technology acting as the key. Technology has a pivotal role in determining states’ geopolitical standing––hence the concept of techno-geopolitics. The ultimate defeat of Japan in the Second World War, for example, proved how technological superiority could significantly alter the outcome of events according to a state’s interest. Technological advancement serves to transform the face of the industry, redefine security measures, and influence the trajectory of social development.[3] That said, by definition, techno-geopolitics refers to the complex interplay between technology and geopolitics, which defines not only global dynamics but also power structure. The struggle for technological dominance is particularly evident in the U.S.-China rivalry, where China has outgrown itself from manufacturers of American-engineered products to a strategic competitor of the U.S. at the research and innovation frontier.[4]

A particular dimension amongst the multi-dimension of the U.S.-China technological race is the procurement of artificial intelligence capabilities. This is contingent on AI’s transformative capability to harbor economic expansion, national defense operation that promotes security and stability, scientific and technological breakthroughs, improved means of global governance, transnational trends, and social change.[5] It must be emphasized, however, that the contemporary AI race is unlike the Cold War nuclear arms race. Weaponries in themselves are ‘command technologies’ that provide a competitive advantage for states that is determinant of the outcome of critical times and the pursuit of national security. Meanwhile, AI falls more fittingly into the category of ‘strategic technologies’ that influence decision-making, automate complex processes, and shape the geopolitical and economic landscape––making it a critical tool for long-term dominance and innovation.

DeepSeek the Meddler for Profits and Exclusivity

For the greater part, the U.S. holds a clear-cut advantage of AI capabilities in comparison to any other countries in the world.[6] This is evidently so, at least in the domain of large language models with ChatGPT at the frontier. The concept of LLMs has existed since the early 20th century, but it was OpenAI’s release of ChatGPT in late 2022 that catalyzed the widespread development and public adoption of generative AI.[7] Appearing seemingly out of the blue, ChatGPT demonstrates that not only could machines ‘learn’ the intricacies of human language but can also ‘mimic’ the complexities and subjectivities of human interactions.[8] With over 400 million weekly active users as of February 2025, ChatGPT’s role in redefining the standards of AI influences more innovation and evolution of AI across a wide range of industries.[9] The success of ChatGPT, marked by its widespread adoption, has fostered a form of urban dependency on the technology, further solidifying America’s influence in the global AI landscape. The emergence of DeepSeek, which has begun to challenge ChatGPT’s dominance, is increasingly regarded as a growing concern for the U.S. This is particularly due to DeepSeek’s operational framework, which disrupts ChatGPT’s business model and has the potential to overturn the economics of the industry and further threaten the techno-geopolitical superiority of the U.S.’s AI. This disruption can be generalized into two means: accessibility and democratization.

DeepSeek is cheaper to develop and even free to access–at least for now, because there has yet to be any information reading for future commodification of the R1 model. There is a prevailing notion that the development of pioneering language models requires massive technical and financial resources. According to Sam Altman, CEO of OpenAI, the training cost of ChatGPT-4o alone was $100 million.[10] This figure does not account for additional expenses such as development, fine-tuning, and operationalization, which are significantly higher. In contrast, the training and inference costs of DeepSeek are said to be less than $6 million, while the model is proven to be operating with excellent accuracy and competitive comparability to its counterpart.[11] This is not to mention that, for users, the pricing range of DeepSeek is far more affordable than ChatGPT by multifolds.[12] The stark contrasts in affordability and accessibility could redirect global reliance away from ChatGPT and other American tech companies–not just among urban users but also among current and future investors. At the same time, since OpenAI’s revenue stream relies on subscriptions and enterprise partnerships, DeepSeek’s low inference costs—without drastically altering its operational efficiency—could significantly weaken OpenAI’s competitive edge and reduce its long-term revenue. From a broader perspective, these shifts and disparities may indirectly weaken U.S. influence over AI governance, standards, and ethical frameworks–and, in turn, strengthen China’s.

Furthermore, unlike ChatGPT, which operates as a proprietary technology, the DeepSeek R1 model is completely open-source.[13] This allows technology companies, organizations, and even well-resourced individuals to freely utilize, adapt, and even monetize the R1 model without limitation.[14] Released under the permissive MIT license, the R1 model ensures broad accessibility and distribution credibility. However, this openness generally comes with both advantages and challenges. On the one hand, Sallam et al. (2025) contend that open-source models hold “the promise of a profound revolution” for innovation, advancement, and collaboration among worldwide developers. The democratization of AI dreams of the barriers of global community enrichment and knowledge that were previously limited by proprietary systems. Inclusivity promotes a transformative shift in problem-solving, particularly to contemporary challenges. On the other hand, Sallam et al. labeled open-source models as “the dawn of doomsday” because while they pave the path for progress and advancement, the lofty ideals of democratization and inclusivity serve as Pandora’s box of unrestrained innovation that may threaten global security and individual agency if used by malicious actors. This is by no means paralleling China with insidious intentions to disrupt global order for its national interest. Rather, it aims to argue from an objective standpoint on the potential misuse and risks of open-source models.

Yet the debate over open-source AI is not just a technical or ethical dilemma; it also carries significant geopolitical weight. As China advances AI accessibility through DeepSeek R1 Models, the US faces rapid erosion in dominant control over AI development. Even without considering the potentiality of misuse as expressed earlier, the decentralization of AI models enables foreign competitors to develop cutting-edge AI in the absence of American technology, which translates to an erosion in its influence over global deployment of AI and the weakening of its technological leverage. Additionally, with open-source models enabling AI to be developed and fine-tuned freely, countries the US has strategically sanctioned from sales of certain tools and technologies, such as NVIDIA’s GPUs, including Russia and Iran, could now proceed with procuring powerful AI models, diminishing the economic and political impact of US sanctions.[15] Ultimately, the democratization of AI capabilities creates a leeway for the incoming of more powerful AI holders that will raise the current AI techno geopolitics from bipolar to multipolar.

  1. Baptista, Eduardo. “What Is DeepSeek and Why Is It Disrupting the AI Sector? | Reuters.” Reuters, January 29, 2025. https://www.reuters.com/technology/artificial-intelligence/what-is-deepseek-why-is-it-disrupting-ai-sector-2025-01-27/. ↑

  2. Kumar, Shubham. “DeepSeek vs ChatGPT: Comparison of Best AI Titans in 2025.” GeeksforGeeks (blog), February 10, 2025. https://www.geeksforgeeks.org/deepseek-vs-chatgpt/. ↑

  3. Sultan, Burree. “Techno-Geopolitics: Unearthing the Future Landscape through the Lens of Technology.” Modern Diplomacy, August 30, 2023. https://moderndiplomacy.eu/2023/08/30/techno-geopolitics-unearthing-the-future-landscape-through-the-lens-of-technology/. ↑

  4. Hass, Ryan. “U.S.-China Relations: The Search for a New Equilibrium.” Brookings Institution, February 2020. https://www.brookings.edu/wp-content/uploads/2020/02/FP_2020026_us_china_relations_hass.pdf. ↑

  5. Valladão, Alfredo G A. “Artificial Intelligence and Political Science,” September 2018. https://www.policycenter.ma/sites/default/files/OCPPC-PP1807_0.pdf. ↑

  6. Sultan, Burree. “Techno-Geopolitics: Unearthing the Future Landscape through the Lens of Technology.” Modern Diplomacy, August 30, 2023. https://moderndiplomacy.eu/2023/08/30/techno-geopolitics-unearthing-the-future-landscape-through-the-lens-of-technology/. ↑

  7. Foote, Keith D. “A Brief History of Large Language Models.” DATAVERSITY, December 28, 2023. https://www.dataversity.net/a-brief-history-of-large-language-models/. ↑

  8. Marr, Bernard. “A Short History Of ChatGPT: How We Got To Where We Are Today.” Forbes, May 19, 2023. https://www.forbes.com/sites/bernardmarr/2023/05/19/a-short-history-of-chatgpt-how-we-got-to-where-we-are-today/. ↑

  9. Carino, Meghan McCarty. “ChatGPT Now Has 400 Million Weekly Users — and Lots of Competition.” Marketplace, February 20, 2025. https://www.marketplace.org/2025/02/20/chatgpt-now-has-400-million-weekly-users-and-a-lot-of-competition/.; Marr, Bernard. “A Short History Of ChatGPT: How We Got To Where We Are Today.” Forbes, May 19, 2023. https://www.forbes.com/sites/bernardmarr/2023/05/19/a-short-history-of-chatgpt-how-we-got-to-where-we-are-today/. ↑

  10. Knight, Will. “OpenAI’s CEO Says the Age of Giant AI Models Is Already Over | WIRED,” April 17, 2023. https://www.wired.com/story/openai-ceo-sam-altman-the-age-of-giant-ai-models-is-already-over/. ↑

  11. The Economic Times. “DeepSeek AI Cost: Chinese AI Startup DeepSeek’s $6 Million vs $250 Billion of Big US Cloud Companies: What about Privacy Concerns? – The Economic Times.” News. The Economic Times, January 28, 2025. https://economictimes.indiatimes.com/news/international/us/chinese-ai-startup-deepseeks-6-million-vs-250-billion-of-big-us-cloud-companies-what-about-privacy-concerns/articleshow/117648917.cms?from=mdr. ↑

  12. Baptista, Eduardo. “What Is DeepSeek and Why Is It Disrupting the AI Sector? | Reuters.” Reuters, January 29, 2025. https://www.reuters.com/technology/artificial-intelligence/what-is-deepseek-why-is-it-disrupting-ai-sector-2025-01-27/. ↑

  13. Kerner, Sean Michael. “DeepSeek Explained: Everything You Need to Know.” WhatIs, February 6, 2025. https://www.techtarget.com/whatis/feature/DeepSeek-explained-Everything-you-need-to-know. ↑

  14. Naughton, John. “DeepSeek: Cheap, Powerful Chinese AI for All. What Could Possibly Go Wrong?” The Guardian, February 1, 2025, sec. Technology. https://www.theguardian.com/technology/2025/feb/01/ai-deepseek-cheap-china-google-apple. ↑

  15. Normile, Dennis. “Chinese Firm’s Faster, Cheaper AI Language Model Makes a Splash.” Magazine. Science, January 15, 2025. https://www.science.org/content/article/chinese-firm-s-faster-cheaper-ai-language-model-makes-splash. ↑

References

Baptista, Eduardo. “What Is DeepSeek and Why Is It Disrupting the AI Sector? | Reuters.” Reuters, January 29, 2025. https://www.reuters.com/technology/artificial-intelligence/what-is-deepseek-why-is-it-disrupting-ai-sector-2025-01-27/.

Carino, Meghan McCarty. “ChatGPT Now Has 400 Million Weekly Users — and Lots of Competition.” Marketplace, February 20, 2025. https://www.marketplace.org/2025/02/20/chatgpt-now-has-400-million-weekly-users-and-a-lot-of-competition/.

Foote, Keith D. “A Brief History of Large Language Models.” DATAVERSITY, December 28, 2023. https://www.dataversity.net/a-brief-history-of-large-language-models/.

Hass, Ryan. “U.S.-China Relations: The Search for a New Equilibrium.” Brookings Institution, February 2020. https://www.brookings.edu/wp-content/uploads/2020/02/FP_2020026_us_china_relations_hass.pdf.

Kerner, Sean Michael. “DeepSeek Explained: Everything You Need to Know.” WhatIs, February 6, 2025. https://www.techtarget.com/whatis/feature/DeepSeek-explained-Everything-you-need-to-know.

Knight, Will. “OpenAI’s CEO Says the Age of Giant AI Models Is Already Over | WIRED,” April 17, 2023. https://www.wired.com/story/openai-ceo-sam-altman-the-age-of-giant-ai-models-is-already-over/.

Kumar, Shubham. “DeepSeek vs ChatGPT: Comparison of Best AI Titans in 2025.” GeeksforGeeks (blog), February 10, 2025. https://www.geeksforgeeks.org/deepseek-vs-chatgpt/.

Marr, Bernard. “A Short History Of ChatGPT: How We Got To Where We Are Today.” Forbes, May 19, 2023. https://www.forbes.com/sites/bernardmarr/2023/05/19/a-short-history-of-chatgpt-how-we-got-to-where-we-are-today/.

Naughton, John. “DeepSeek: Cheap, Powerful Chinese AI for All. What Could Possibly Go Wrong?” The Guardian, February 1, 2025, sec. Technology. https://www.theguardian.com/technology/2025/feb/01/ai-deepseek-cheap-china-google-apple.

Normile, Dennis. “Chinese Firm’s Faster, Cheaper AI Language Model Makes a Splash.” Magazine. Science, January 15, 2025. https://www.science.org/content/article/chinese-firm-s-faster-cheaper-ai-language-model-makes-splash.

Sallam, Malik, Kholoud Al-Mahzoum, Mohammed Sallam, and Maad M. Mijwil. “DeepSeek: Is It the End of Generative AI Monopoly or the Mark of the Impending Doomsday?” Mesopotamian Journal of Big Data 2025 (January 30, 2025): 26–34. https://doi.org/10.58496/MJBD/2025/002.

Sultan, Burree. “Techno-Geopolitics: Unearthing the Future Landscape through the Lens of Technology.” Modern Diplomacy, August 30, 2023. https://moderndiplomacy.eu/2023/08/30/techno-geopolitics-unearthing-the-future-landscape-through-the-lens-of-technology/.

The Economic Times. “DeepSeek AI Cost: Chinese AI Startup DeepSeek’s $6 Million vs $250 Billion of Big US Cloud Companies: What about Privacy Concerns? – The Economic Times.” News. The Economic Times, January 28, 2025. https://economictimes.indiatimes.com/news/international/us/chinese-ai-startup-deepseeks-6-million-vs-250-billion-of-big-us-cloud-companies-what-about-privacy-concerns/articleshow/117648917.cms?from=mdr.

Valladão, Alfredo G A. “Artificial Intelligence and Political Science,” September 2018. https://www.policycenter.ma/sites/default/files/OCPPC-PP1807_0.pdf.

Wong, Pak Nung. Techno-Geopolitics: U.S.-China Tech War and the Practice of Digital Statecraft. 1st ed. London: Routledge India, 2021. https://doi.org/10.4324/9781003047100.