ANALYSIS: China and AI — Dominance That No Longer Asks for Permission

A Model Built Under Sanctions

The story of DeepSeek alone is a lesson in humility for those who believed that U.S. sanctions on microchips would be enough to decisively slow down Chinese AI. Since 2022, Washington has gradually banned the export to China of Nvidia’s advanced GPUs—notably the A100 and H100 series—the processors that form the backbone of training large language models. The goal was explicit: to deprive China of the computational tools needed to compete in the race for large AI models.

DeepSeek has demonstrated that this containment strategy has fundamental limitations. By developing radically more efficient optimization techniques—notably Mixture of Experts (MoE) and innovative model compression approaches—Liang Wenfeng’s team produced a model competitive with GPT-4 and Claude at a fraction of the computational cost. Where OpenAI would have spent hundreds of millions of dollars on infrastructure, DeepSeek reportedly achieved comparable performance for about $6 million—a figure that remains subject to debate, but whose order of magnitude is telling.

Innovation Under Constraints as a Catalyst

There is a cruel irony in the fact that U.S. technology sanctions may, paradoxically, have spurred Chinese innovation rather than stifling it. Deprived of access to the most advanced chips, Chinese engineering teams were forced to rethink their computational approaches from the ground up. They developed more efficient algorithms, more frugal model architectures, and lighter inference techniques. Once available, these innovations benefit more than just China—they redefine the global state of the art.

Strategists in Washington have made a classic mistake: believing that superiority in resources guarantees technological victory. Yet the history of innovation is full of players who have turned their constraints into sustainable competitive advantages. DeepSeek is not an anomaly. It may be the vanguard of a new paradigm.

Sector-Wide Reactions

The shockwaves from DeepSeek were not limited to the stock markets. They triggered a series of chain reactions throughout the global technology ecosystem. Microsoft, Google, Meta, and Amazon have all accelerated their announcements of AI investments. The U.S. Congress held emergency hearings. The European Union has reopened debates on its technological sovereignty. And in research labs around the world, entire teams have begun studying the Hangzhou team’s optimization techniques to understand what they had overlooked.

The U.S. Embargo and Its Structural Flaws

The U.S. policy of controlling exports of advanced semiconductors to China represents one of the most complex fronts in the Sino-American technological rivalry. Initiated under the Biden administration and strengthened on several occasions, this policy aims to limit China’s access to the high-performance computing chips that power large AI models. It involves not only direct restrictions on U.S. exports but also pressure on allies—notably the Netherlands, home to ASML, the world’s sole manufacturer of EUV lithography machines essential for the most advanced process nodes—to align their trade policies.

These efforts have had tangible effects: China has lost access to the most advanced manufacturing technologies, particularly nodes below 7 nanometers. But they face structural limitations that are difficult to overcome. The global market for less advanced semiconductors remains largely accessible. Alternative trade routes exist via third countries. And above all, China has invested heavily in developing its own domestic chip manufacturing capabilities, centered on its national champion SMIC (Semiconductor Manufacturing International Corporation) and an ecosystem of subcontractors.

SMIC and the Substitution Strategy

In 2023, the world was stunned to learn that SMIC had succeeded in producing 7-nanometer chips for the Kirin 9000S processor in the Huawei Mate 60 Pro smartphone—without access to ASML’s EUV machines, using multi-pass DUV lithography techniques. Western experts had deemed this technically unlikely. Yet it was a reality. It’s not that China has achieved full parity with TSMC or Samsung—that would be inaccurate. But it has demonstrated an ability to bridge technological gaps once thought insurmountable, using available resources and creative approaches.

The lesson from semiconductors is that the gap is narrowing faster than expected, through paths no one had anticipated. Every time the West has set a technological limit that China could not cross, China has found a way to cross it by other means. At some point, we must recognize that this is a pattern, not a series of coincidences.

The 2030 Outlook for Chinese Semiconductors

Chinese investment in the domestic semiconductor industry is on an unprecedented scale. The National Integrated Circuit Fund, commonly known as the “Big Fund,” has mobilized tens of billions of dollars in several phases since 2014. Provinces and municipalities have added their own funding. Hundreds of startups specializing in chip design have emerged. University programs have been created from scratch to train tens of thousands of specialized engineers. The result is an industry that is still incomplete and still dependent on imports in certain critical segments, but one that is advancing at a pace that sanctions alone will not be enough to halt.

A Closed Digital Ecosystem as a Strategic Asset

One of the least understood aspects of China’s potential superiority in artificial intelligence lies in the very structure of its digital space. The Chinese internet, often portrayed negatively in terms of censorship and the Great Firewall, also constitutes a closed data ecosystem of immense strategic value. Companies operating in this space—Baidu, Alibaba, Tencent, ByteDance, Meituan, and Didi—capture and centralize behavioral, transactional, linguistic, and biometric data on a scale that even the American GAFAM companies struggle to match in terms of density and granularity.

This data fuels AI models. The quality of a large language model, a facial recognition system, a recommendation engine, or a medical diagnostic algorithm depends directly on the richness of the data on which it was trained. On this front, China holds advantages that cannot be replicated in the short term by Western democracies, which face legitimate but real constraints regarding the protection of personal data.

Industrial Data as a New Battleground

Beyond personal data, China is also capitalizing on its industrial and manufacturing data. As the world’s leading producer in dozens of sectors—from electronics to automobiles, from renewable energy to fine chemicals—China generates colossal volumes of data on production, quality, logistics, and performance. This data, combined with industrial AI systems, creates cycles of continuous optimization that further strengthen manufacturing competitiveness. AI fuels industry, and industry fuels AI. This virtuous cycle is one of the most powerful—and least visible—drivers of China’s technological rise.

There is something almost metaphysical about this accumulation. China is not merely building more powerful AI models. It is building the structural conditions that will ensure its models remain superior in the long term. Data is not a finite resource—it accumulates, enriches itself, and grows in complexity. And the advantage it provides is, by nature, cumulative.

Data Regulation as a Tool of Sovereignty

Beijing has also used data regulation as an instrument of industrial policy. The Data Security Law (2021) and the Personal Information Protection Law (2021) have imposed strict rules on the transfer of data out of China, forcing foreign companies operating within the country to localize their data. This policy serves a dual purpose: to preserve Chinese data as an exclusive national resource, and to create asymmetries in the availability of data for model training—to the detriment of foreign players seeking to compete in the Chinese market.

The Educational System as Strategic Infrastructure

Behind every line of code, every algorithmic breakthrough, and every scientific publication, there are engineers. And on this front, China has invested with remarkable intensity. The number of Chinese graduates in science, technology, engineering, and mathematics (STEM) far exceeds that of any other country. In AI specifically, programs of excellence have been established at dozens of universities—Tsinghua, Peking, Zhejiang, Fudan, USTC—with massive public funding and aggressive international recruitment.

The volume is impressive. So is the quality. International AI research rankings consistently show Chinese teams among the most prolific and most-cited in the fields of deep learning, computer vision, natural language processing, and robotics. The era when Chinese research was limited to replicating Western advances is over. It now leads the way on certain fronts.

The Diaspora as a Bridge and a Challenge

China’s relationship with its scientific diaspora is one of the most complex and strategically important aspects of this technological competition. Hundreds of thousands of researchers and engineers of Chinese origin work in Western laboratories and companies—at Google, Meta, Microsoft, MIT, Stanford, and CMU. They contribute to the advancement of global research, and their knowledge circulates—through academic publications, international conferences, and sometimes through more direct channels.

The issue of the diaspora holds up a mirror to our contradictions. For decades, the West has benefited from Chinese talent trained at its universities. Today, it finds itself debating whether that same talent constitutes a vehicle for undesirable technology transfer. The answer to this question says as much about our own fears as it does about the intentions of the people involved.

The Brain Gain and the Thousand Talents Program

At the same time, China has rolled out aggressive programs to entice the return of its expatriate researchers. The Thousand Talents Program—despite the controversies it has sparked in the United States and the FBI investigations that led to its unofficial renaming—has helped repatriate hundreds of top-tier researchers from Western universities and companies. These programs offer competitive financial packages, state-of-the-art laboratories, substantial research resources, and, increasingly, the opportunity to work on cutting-edge technological challenges. For many, the calculus has changed.

The Doctrine of Civil-Military Integration

One of the most concerning aspects for Western strategists regarding the rise of Chinese AI is the official doctrine of civil-military integration (junmin ronghe). This policy, elevated to a national priority under Xi Jinping, aims to eliminate the boundary between civilian technological development and military applications. In practical terms, this means that advances made by companies such as Baidu, Alibaba, and Huawei in fields like facial recognition, autonomous navigation, and natural language processing can be directly harnessed for the benefit of the People’s Liberation Army.

This interpenetration between the civilian and military sectors is difficult to replicate in liberal democracies, where institutional and legal safeguards separate the two spheres. It constitutes a competitive advantage in the military transformation that China has been pursuing for the past decade. Autonomous drones, mass surveillance systems, offensive cybersecurity, electronic warfare, and AI-optimized military logistics—all these areas benefit from Beijing’s ability to simultaneously mobilize resources from across its entire economy.

AI in Weapons Systems and Surveillance

The practical applications of Chinese military AI are now evident. Swarms of autonomous drones, AI-guided missile defense systems, and biometric surveillance and identification platforms deployed in Xinjiang and other regions—and exported to authoritarian partner governments as part of the Digital Silk Road—demonstrate an operational capability that extends far beyond research laboratories. The question is no longer whether China can develop these systems. It has already done so. The question now is their global proliferation and their implications for the international security order.

There is a perverse form of intellectual comfort in the Western tendency to separate “civilian” and “military” Chinese AI, as if the same algorithms, the same data, and the same engineers did not serve both. Civil-military convergence is not a political metaphor. It is a real institutional architecture, and its strategic implications are catching up with us faster than our ethical debates.

Technology Export as an Instrument of Power

Beyond its own applications, China uses its AI technologies as a tool for projecting global power. AI-based urban surveillance systems developed by companies such as Hikvision, Dahua, and Huawei have been deployed in more than 80 countries, often as part of contracts financed by Chinese loans. This proliferation creates technological dependencies, data flows to Chinese servers, and ties of strategic interdependence that serve Beijing’s geopolitical interests far beyond mere commercial transactions.

The United States: Between Sanctions and Massive Investments

Faced with the rise of Chinese AI, the United States has deployed two types of responses: containment through sanctions and revitalization through public investment. The CHIPS and Science Act of 2022, which allocates $52 billion to bring semiconductor production back to the United States, and the Creating Helpful Incentives to Produce Semiconductors Act represent an effort to rebuild domestic capacity in a sector that had been largely outsourced to Asia. The first results are coming in: TSMC is building factories in Arizona, Intel is investing heavily in new fabs, and Samsung is setting up shop in Texas.

But these investments take time. Building a state-of-the-art semiconductor factory takes 3 to 5 years. Training the engineers to run it takes even longer. And in the meantime, China continues to advance its own capabilities. There is a real risk of a race in which the West invests heavily to catch up, only to see the gap widen simultaneously on other fronts.

Europe: Between Sovereign Ambition and Industrial Reality

The situation in Europe is even more concerning. Despite bold rhetoric about digital sovereignty and strategic autonomy, the European Union has no equivalent to OpenAI, Anthropic, Google DeepMind, or DeepSeek. European AI startups—Mistral in France, Aleph Alpha in Germany—show promise but are incomparably smaller in terms of resources and scale. The European AI Act, adopted in 2024, is the world’s most comprehensive regulatory framework—but it regulates an industry in which Europe does not hold a dominant position and risks stifling the very innovation it seeks to foster.

Europe has produced the world’s most sophisticated regulation for an industry in which it is a minority player. This is a remarkable achievement in regulatory governance—and potentially a major competitive disadvantage. Regulating what others are building is like being the referee of a game you’re not playing.

Technological Alliances as a Collective Response

Faced with the scale of the challenge, democracies are beginning to develop collective responses. The AUKUS framework now includes a substantial technological component. The Quad (United States, Japan, Australia, India) has established working groups on AI and semiconductors. G7 summits regularly address coordination on AI governance. And bilateral agreements are proliferating among countries that share common concerns about China’s access to strategic technologies. These efforts are real but suffer from insufficient coordination and a pace of implementation that remains slower than Beijing’s.

Medicine as a Testing Ground for Mass Experimentation

Digital health is one of the areas where Chinese AI has made its most spectacular and tangible advances. AI-assisted medical diagnostic systems developed by companies such as Infervision, Deepwise, and Ping An Health are being deployed in hundreds of Chinese hospitals, processing millions of cases and accumulating training data that their Western competitors can only envy. The reported performance in detecting cancers, radiological abnormalities, or retinal pathologies rivals—and sometimes surpasses—that of the most renowned American systems.

Scale matters. A medical AI system deployed in a hospital network serving 50,000 patients per day learns incomparably faster than a system tested in a clinical trial involving just a few hundred cases. China has access to this massive flow of real-world medical data, within a regulatory framework that facilitates its use for training models. The result is an accelerated learning curve that produces increasingly high-performing systems, faster and faster.

Energy and Climate as Key Deployment Areas

The energy transition is another sector where the convergence of Chinese industrial leadership and AI capabilities creates powerful synergies. China is the world’s leading producer of solar panels, lithium-ion batteries, and electric vehicles. It also operates the world’s largest fleet of offshore wind turbines. Optimizing these complex energy systems—including production forecasting, balancing smart grids, and predictive maintenance—is a natural application area for industrial AI, and China is accumulating operational data and deployment experience at an unparalleled rate.

There is an almost dizzying symbiosis between China’s industrial dominance in clean technologies and its growing capabilities in AI. Every wind turbine, every battery, every electric vehicle that rolls off a Chinese assembly line generates data that feeds AI systems, which in turn optimize the next generation of wind turbines, batteries, and vehicles. It is a technological flywheel whose momentum grows with every rotation.

Robotics and Industry 4.0

Industrial robotics is perhaps the area where the implications of Chinese AI will be most transformative for the global economy. China is already the world’s largest market for industrial robots, with more than 290,000 units installed in 2023. It is also on track to become one of the leading producers, with companies such as ESTUN, SIASUN, and Inovance rapidly moving upmarket. The integration of AI into these robotic systems—enabling real-time adaptation, reinforcement learning, and human-robot collaboration—is transforming Chinese factories into laboratories for advanced automation, with the lessons learned subsequently exported around the world.

The Issue of Global Technological Dependence

As Chinese AI technologies spread globally—through consumer apps like TikTok and DeepSeek, through infrastructure systems like those from Huawei, and through health and education platforms deployed in developing countries—the issue of the technological dependencies created by this spread has become increasingly acute. The U.S. decision to restrict TikTok has highlighted a fundamental concern: what happens when hundreds of millions of people entrust their data, behaviors, and preferences to platforms that may be under the influence of a foreign government?

This question is not limited to the United States. It arises wherever Chinese technologies penetrate critical markets. In Kenya, Ethiopia, Pakistan, Indonesia—and in dozens of emerging economies that have adopted Chinese digital infrastructure because it was cheaper and more readily available than Western alternatives—the issue of data sovereignty and technological dependence is becoming increasingly urgent.

AI Governance: Between Regulation and Competition

Global AI governance is one of the most complex challenges of our time, precisely because the main players in this technology—the United States and China—have divergent interests regarding how it should be regulated. China has developed its own regulatory frameworks for AI, particularly for algorithmic recommendation systems and AI-generated content, which serve both internal control objectives and as a model for export to partner countries. These Chinese AI governance standards compete with Western approaches to influence international norms—notably at the ITU (International Telecommunication Union) and in other technical standardization bodies.

The battle over technical standards is invisible to the general public but crucial to the architecture of tomorrow’s digital world. Whoever sets the standards sets the rules of the game. And in this battle, too, China has organized itself with a determination and consistency that liberal democracies struggle to match—precisely because their pluralism makes coordination slower, more difficult, and more costly.

Human Rights and the Ethics of Surveillance AI

It is impossible to discuss the rise of Chinese AI without addressing its most troubling applications. The facial recognition, mass surveillance, and social credit technologies deployed in China represent the most extensive application of AI for population control ever seen in human history. The total surveillance of Xinjiang—where more than one million members of the Uyghur minority have been interned in re-education camps identified and targeted in part through AI systems—is an extreme and well-documented example of the potential abuses of these technologies.

Areas Where China Already Leads

With China’s official goal of global AI dominance by 2030 on the horizon, what is the progress so far? On some fronts, dominance is already a reality. In terms of the volume of academic publications, China has ranked first in the world since 2020. In facial recognition, its systems are among the most advanced in the world. In the field of machine translation to and from Mandarin, it is unquestionably in the lead. For AI applied to e-commerce, logistics, and digital payments, its platforms operate at scales and with levels of sophistication that their Western counterparts cannot match.

In the field of autonomous vehicles, competition is fierce, but China has significant advantages: a more permissive regulatory framework for testing, high volumes of urban traffic that accelerate system learning, and manufacturers such as BYD and SAIC that are integrating these technologies on a massive scale into their new models. And in the realm of large language models, DeepSeek has demonstrated that China can not only rival the best Western models but potentially surpass them in efficiency.

Areas Where a Gap Still Exists

It would be inaccurate to paint a picture of total dominance. Substantial gaps persist. In fundamental research in theoretical AI, U.S. and British institutions maintain a lead in certain subfields. In cutting-edge semiconductor manufacturing technologies, China remains about one technological generation behind TSMC and Samsung. In terms of the open and collaborative ecosystems that have characterized much of Western AI innovation, the culture of sharing and publishing is more developed in U.S. laboratories—even though this trend is reversing as geopolitical tensions escalate.

Recognizing China’s strengths in AI does not mean losing hope in the West’s ability to rise to the challenge. But we must first see the situation clearly for what it is, not for what we would like it to be. Years of technological complacency toward China have come at a high cost in terms of strategic preparedness. It would be costly to continue down this path.

Scenarios for 2030

Several scenarios are emerging for the year 2030. In the first, China achieves its goal of global leadership in AI, dominating fundamental research, industrial applications, and the international dissemination of its technologies and standards. In the second, a bipolar competition takes hold on a lasting basis, with two parallel technological ecosystems—one centered on the United States and its allies, the other on China and its partners—advancing at comparable speeds but according to distinct paradigms. In the third scenario, Western democracies—spurred by the awareness raised by DeepSeek and other warning signs—mobilize their resources in a sufficiently coordinated manner to maintain or regain significant technological leadership.

What the Numbers Really Say

The global race for artificial intelligence is not over. In many ways, it has barely begun. The models we consider today to be the pinnacle of technology will be outdated in two years. The applications that seem revolutionary to us in 2025 will be commonplace by 2030. In this accelerating race, what matters is not current position but trajectory: who is rising, who is stagnating, and who is falling? On this metric, the answer is clear. China is rising. And it is rising fast, consistently, on many fronts simultaneously.

The numbers speak for themselves. More than 40% of global AI publications today are of Chinese origin. The country trains more than 2 million engineers in STEM disciplines each year. It invests tens of billions of dollars annually in research and technological development. It deploys these technologies on a domestic scale that creates unparalleled learning advantages. And it possesses strategic consistency—a state that says what it does and does what it says—which is both its greatest asset and, for those living under its authority, its greatest threat.

What the West Must Decide

The response to the rise of Chinese AI cannot be reduced to export sanctions and rhetoric about democratic values. It requires an industrial, educational, and institutional mobilization on a scale comparable to that which democracies have deployed during other moments of existential competition—the Cold War, the space race, and postwar reconstruction. It also requires an uncompromising clarity about the inherent flaws in liberal societies’ ability to coordinate and execute long-term technological strategies.

The history of technological dominance is not a foregone conclusion. No hegemony is permanent, and China faces its own considerable challenges—demographic, economic, and institutional. But the time for complacency is clearly over. What is being built in Beijing, Hangzhou, Shenzhen, and in Chinese universities is no longer merely a catch-up effort. It is an alternative. And this alternative deserves to be taken seriously—not with paralyzing fear, but with the courageous clarity of those who choose to understand before acting.

A Reality That No Longer Asks for Permission

There is something fundamentally new about the nature of this competition. China no longer seeks Western validation of its technological advances. It no longer asks permission to innovate. It makes no apologies for its ambitions. It builds, it deploys, it exports, it sets standards. DeepSeek was not an isolated coup—it was a demonstration of technological maturity by a civilization that has decided to play in the big leagues and that proves, week after week, that it has what it takes. The West can choose to view this as a threat to be contained, or as a challenge that could—if taken seriously—lift up all of humanity. This choice of perspective is not insignificant. It will define the nature of the decades to come.

Signed, Jacques Pj Provost