17 March 2025
CO25056 | DeepSeek: A Seed for Technological Equity and a Multipolar World
SYNOPSIS
In 2025, Chinese AI firm DeepSeek disrupted US tech hegemony by launching open-source models R1/V3, rivalling top-tier AI with cost-efficient innovation. The widespread adoption of the app challenged US technology dominance, with significant implications for international politics.
COMMENTARY
In early 2025, Chinese AI firm DeepSeek launched its R1 and V3 models, swiftly topping app store download charts in China and the US, directly challenging the dominance of American AI giants and stunning the global tech community. From February 24 to 28, DeepSeek held an “Open Source Week”, releasing a core codebase daily, showcasing its genuine commitment to open-source principles and signalling its push for global AI equity. This move once again captured the world’s attention.
Shattering the US AI Monopoly
American AI titans once fortified their technological monopoly with a “capital moat” – OpenAI poured billions into training GPT-4, Google amassed trillion-token datasets, and Nvidia controlled 90 percent of the global high-end GPU supply. Yet, DeepSeek-V3, built with just US$5.58 million and mid-tier H800 GPUs, achieved performance rivalling top-tier models, shattering the myth of “compute supremacy” in AI.
Harvard AI expert Louis Tompros remarked, “DeepSeek demonstrates that algorithmic innovation can bridge the hardware gap, posing an unprecedented challenge to US technological hegemony”. Under this pressure, OpenAI, Google, Meta, Microsoft, and Nvidia – while clinging to proprietary core tech – were forced to open-source some AI tools and slash service fees.
On January 27, Nvidia’s market cap plummeted by US$589 billion, reflecting shaken confidence in the hardware dependency of US AI giants. The International Data Corporation (IDC) predicts that by 2026, the global open-source AI market share will rise from 15 percent in 2024 to 30 percent, with DeepSeek playing a pivotal role. This suggests that America’s global AI monopoly is facing a serious reckoning.
Rapid Embrace from Global Majority
Following the debut of DeepSeek-R1 and V3, only Italy imposed a nationwide ban, while Australia, South Korea, and the US saw partial or full government restrictions. In stark contrast, most of the world embraced DeepSeek. The EU viewed it as a chance to achieve “digital sovereignty”, with European Commission President Ursula von der Leyen stating, “DeepSeek’s inclusivity aligns with the EU’s ethical AI strategy”.
German industry is exploring DeepSeek’s MoE (Mixture-of-Experts) architecture to optimise supply chains, while its quality inspection systems have integrated the technology. France’s Mistral AI partnered with DeepSeek to advance open-source models, Japan’s SoftBank eyed its potential for robotics, and Global South nations showed eager enthusiasm.
Within 18 days of the R1 release, India, with 15.6 percent of downloads, emerged as the largest growth market. IT Minister Ashwini Vaishnaw confirmed deployment on local servers, with firms like TCS, Infosys, and Shorthills AI seizing the innovation opportunity.
Brazil, Vietnam, and Nigeria are testing DeepSeek in agriculture and education, underscoring its open-source appeal. Even in the US, Meta and Snowflake plan to adopt DeepSeek tech. Perplexity and Grok have already integrated its models, and OpenAI dropped earlier lawsuit threats, shifting to learn from DeepSeek’s innovations.
This widespread acceptance highlights how DeepSeek is accelerating global AI equity and softening the US-China AI decoupling trend.
Igniting a Global Innovation Renaissance
More strikingly, DeepSeek’s breakthrough – matching US top-tier models with low costs, mid-range chips, and homegrown talent – shattered the notion that China can only trail the US in tech. It proved that China possesses comparable talent, capital, and market potential, lacking only the confidence and courage to innovate.
Companies like DeepSeek and Unitree Robotics have ignited that spark, fueling a new wave of Chinese tech ambition. This vitality is rippling globally, inspiring the EU and Global South nations to pursue independent AI development outside the US-China binary.
A fresh surge of collaborative innovation has emerged in the global AI community. During DeepSeek’s Open Source Week, over 100,000 developers downloaded its code, and GitHub “stars” soared to 150,000, reflecting the scale of this cooperative momentum.
Fostering a Political Will for AI Equity
DeepSeek’s rise coincides with a growing global political consensus on AI equity. On February 11, 2025, 61 nations signed the Paris AI Action Summit Joint Declaration, endorsing openness, inclusivity, ethics, sustainability, and global dialogue as core principles while rejecting AI hegemony.
The Munich Security Report 2025: Multipolarity identifies AI as a driver of multipolarity, urging a framework rooted in the UN Charter and Universal Declaration of Human Rights to ensure tech accessibility. DeepSeek’s emergence has turbocharged this collective will. Founder Liang Wenfeng said, “AI should not be a sceptre of power but a public tool for human civilisation”, a vision transcending mere commercial goals. Scholar Joseph Nye noted, “AI equity is now a multipolar consensus; the US risks losing allies if it resists this tide”.
Yet, the US has shifted its China AI policy from tech restriction to full containment, clinging to hegemonic ambitions. This approach is set to backfire. Not only will it spur Chinese innovation, but it will also isolate the US in an AI equity era.
The latest Munich Security Report shows trust in US AI policy among non-G7 nations dropping from 62 percent in 2023 to 48 percent in 2025, signalling that America’s monopolistic image is driving countries toward open-source collaboration. AI equity and multipolarity are becoming unstoppable.
Shaping a Multipolar World
Framed in the US media’s typical US-China rivalry lens, the author of this commentary asked Grok 3 to forecast the global AI trajectory. It predicted a bipolar US-China pattern. But when he posed, “What if China prioritises cooperative AI development with most nations over confronting the US?” Grok 3 replied: US-China competition would soften and become indirect, shifting the global AI landscape from bipolar to multipolar.
The EU and Global South would emerge as key variables, fostering distinct ecosystems. While the US and China remain frontrunners, their influence will noticeably disperse. Grok 3 affirmed DeepSeek’s role, stating, “DeepSeek’s open-source model could be the seed of tech equity, birthing a decentralised AI world”.
In February 2025, China-ASEAN AI projects surged 40 percent, and the EU’s Gaia-X began integrating open-source tech – early signs of a multipolar AI future that will inevitably reshape international politics.
Thus, for any nation, abandoning narrow self-interest, avoiding cutthroat competition, and embracing AI equity and multipolarity while focusing on managing the security risks of this shift is the right path to co-create a fair, multipolar AI era.
About the Author
Dr Qi Dapeng is a professor at the Center for Strategic Studies, National Defense University (NDU) of the People’s Liberation Army (PLA), Beijing, China. He was a visiting Senior Research Fellow at the S. Rajaratnam School of International Studies (RSIS), Nanyang Technological University (NTU), Singapore, in February 2025.
SYNOPSIS
In 2025, Chinese AI firm DeepSeek disrupted US tech hegemony by launching open-source models R1/V3, rivalling top-tier AI with cost-efficient innovation. The widespread adoption of the app challenged US technology dominance, with significant implications for international politics.
COMMENTARY
In early 2025, Chinese AI firm DeepSeek launched its R1 and V3 models, swiftly topping app store download charts in China and the US, directly challenging the dominance of American AI giants and stunning the global tech community. From February 24 to 28, DeepSeek held an “Open Source Week”, releasing a core codebase daily, showcasing its genuine commitment to open-source principles and signalling its push for global AI equity. This move once again captured the world’s attention.
Shattering the US AI Monopoly
American AI titans once fortified their technological monopoly with a “capital moat” – OpenAI poured billions into training GPT-4, Google amassed trillion-token datasets, and Nvidia controlled 90 percent of the global high-end GPU supply. Yet, DeepSeek-V3, built with just US$5.58 million and mid-tier H800 GPUs, achieved performance rivalling top-tier models, shattering the myth of “compute supremacy” in AI.
Harvard AI expert Louis Tompros remarked, “DeepSeek demonstrates that algorithmic innovation can bridge the hardware gap, posing an unprecedented challenge to US technological hegemony”. Under this pressure, OpenAI, Google, Meta, Microsoft, and Nvidia – while clinging to proprietary core tech – were forced to open-source some AI tools and slash service fees.
On January 27, Nvidia’s market cap plummeted by US$589 billion, reflecting shaken confidence in the hardware dependency of US AI giants. The International Data Corporation (IDC) predicts that by 2026, the global open-source AI market share will rise from 15 percent in 2024 to 30 percent, with DeepSeek playing a pivotal role. This suggests that America’s global AI monopoly is facing a serious reckoning.
Rapid Embrace from Global Majority
Following the debut of DeepSeek-R1 and V3, only Italy imposed a nationwide ban, while Australia, South Korea, and the US saw partial or full government restrictions. In stark contrast, most of the world embraced DeepSeek. The EU viewed it as a chance to achieve “digital sovereignty”, with European Commission President Ursula von der Leyen stating, “DeepSeek’s inclusivity aligns with the EU’s ethical AI strategy”.
German industry is exploring DeepSeek’s MoE (Mixture-of-Experts) architecture to optimise supply chains, while its quality inspection systems have integrated the technology. France’s Mistral AI partnered with DeepSeek to advance open-source models, Japan’s SoftBank eyed its potential for robotics, and Global South nations showed eager enthusiasm.
Within 18 days of the R1 release, India, with 15.6 percent of downloads, emerged as the largest growth market. IT Minister Ashwini Vaishnaw confirmed deployment on local servers, with firms like TCS, Infosys, and Shorthills AI seizing the innovation opportunity.
Brazil, Vietnam, and Nigeria are testing DeepSeek in agriculture and education, underscoring its open-source appeal. Even in the US, Meta and Snowflake plan to adopt DeepSeek tech. Perplexity and Grok have already integrated its models, and OpenAI dropped earlier lawsuit threats, shifting to learn from DeepSeek’s innovations.
This widespread acceptance highlights how DeepSeek is accelerating global AI equity and softening the US-China AI decoupling trend.
Igniting a Global Innovation Renaissance
More strikingly, DeepSeek’s breakthrough – matching US top-tier models with low costs, mid-range chips, and homegrown talent – shattered the notion that China can only trail the US in tech. It proved that China possesses comparable talent, capital, and market potential, lacking only the confidence and courage to innovate.
Companies like DeepSeek and Unitree Robotics have ignited that spark, fueling a new wave of Chinese tech ambition. This vitality is rippling globally, inspiring the EU and Global South nations to pursue independent AI development outside the US-China binary.
A fresh surge of collaborative innovation has emerged in the global AI community. During DeepSeek’s Open Source Week, over 100,000 developers downloaded its code, and GitHub “stars” soared to 150,000, reflecting the scale of this cooperative momentum.
Fostering a Political Will for AI Equity
DeepSeek’s rise coincides with a growing global political consensus on AI equity. On February 11, 2025, 61 nations signed the Paris AI Action Summit Joint Declaration, endorsing openness, inclusivity, ethics, sustainability, and global dialogue as core principles while rejecting AI hegemony.
The Munich Security Report 2025: Multipolarity identifies AI as a driver of multipolarity, urging a framework rooted in the UN Charter and Universal Declaration of Human Rights to ensure tech accessibility. DeepSeek’s emergence has turbocharged this collective will. Founder Liang Wenfeng said, “AI should not be a sceptre of power but a public tool for human civilisation”, a vision transcending mere commercial goals. Scholar Joseph Nye noted, “AI equity is now a multipolar consensus; the US risks losing allies if it resists this tide”.
Yet, the US has shifted its China AI policy from tech restriction to full containment, clinging to hegemonic ambitions. This approach is set to backfire. Not only will it spur Chinese innovation, but it will also isolate the US in an AI equity era.
The latest Munich Security Report shows trust in US AI policy among non-G7 nations dropping from 62 percent in 2023 to 48 percent in 2025, signalling that America’s monopolistic image is driving countries toward open-source collaboration. AI equity and multipolarity are becoming unstoppable.
Shaping a Multipolar World
Framed in the US media’s typical US-China rivalry lens, the author of this commentary asked Grok 3 to forecast the global AI trajectory. It predicted a bipolar US-China pattern. But when he posed, “What if China prioritises cooperative AI development with most nations over confronting the US?” Grok 3 replied: US-China competition would soften and become indirect, shifting the global AI landscape from bipolar to multipolar.
The EU and Global South would emerge as key variables, fostering distinct ecosystems. While the US and China remain frontrunners, their influence will noticeably disperse. Grok 3 affirmed DeepSeek’s role, stating, “DeepSeek’s open-source model could be the seed of tech equity, birthing a decentralised AI world”.
In February 2025, China-ASEAN AI projects surged 40 percent, and the EU’s Gaia-X began integrating open-source tech – early signs of a multipolar AI future that will inevitably reshape international politics.
Thus, for any nation, abandoning narrow self-interest, avoiding cutthroat competition, and embracing AI equity and multipolarity while focusing on managing the security risks of this shift is the right path to co-create a fair, multipolar AI era.
About the Author
Dr Qi Dapeng is a professor at the Center for Strategic Studies, National Defense University (NDU) of the People’s Liberation Army (PLA), Beijing, China. He was a visiting Senior Research Fellow at the S. Rajaratnam School of International Studies (RSIS), Nanyang Technological University (NTU), Singapore, in February 2025.
