C

hina has designated artificial intelligence as a strategic technology for enhancing its economic competitiveness, bolstering military capabilities, and strengthening the Chinese Communist Party's (CCP) governance and mass surveillance systems. Party and state leader Xi Jinping views AI through the lenses of national security and geopolitics. “Whoever can grasp the opportunities of new economic development such as big data and artificial intelligence will have the pulse of our times,” Xi told BRICS members in June 2022. He frequently describes the current techno-industrial revolution, driven by AI, as an unprecedented chance for China to “overtake on the curve”—surpass advanced economies and restore its position as a global power.

 

With strong government support and close ties between industry and state-backed academic labs, China’s AI ecosystem has made impressive strides. This is evidenced by the increasing volume and quality of its research, a vast and rapidly expanding talent pool, significant state and private-sector investment, and the swift adoption of AI in specific sectors. The release of ChatGPT in November 2022 caused a frenzy in China’s generative AI sector, with at least 50 companies developing large language models. These range from major tech firms like Baidu and Huawei to startups like Zhipu AI and 01.AI. Although generally considered to be one to two years behind their US counterparts, some of these models perform quite well on various benchmarks.

Below, we examine three key challenges China faces in leveraging generative AI and other AI innovations to transform its economy: adopting AI to boost productivity, managing the sheer energy consumption of AI systems, and coping with restricted access to foreign semiconductor technology. The analysis concludes by highlighting some of the implications for Europe.

 

 

 

 

 

Enhancing productivity through AI is easier said than done

 

Since coming to power in 2012, Xi Jinping has prioritized national security, resilience, and geopolitical goals over economic growth, efficiency, and market-oriented reforms. He envisions a state-driven economy where the party-state directs resources towards technologies and industries it considers strategic. For Xi, AI is a key driver of “new quality productivity” and the modernization of China’s manufacturing base, which he sees as the main/prime source of China’s competitive advantage. As he put it, “AI’s ‘head goose’ effect will create spillover benefits and drive innovation in other sectors.”

China’s priority is to apply AI across the “real economy,” boosting productivity in traditional sectors like manufacturing. The country is grappling with a prolonged economic downturn, an aging population, and slowing productivity growth. In response, the government is focusing on the technology- and innovation-driven upgrading of China’s industrial base—an agenda that took center stage at the CCP Central Committee’s delayed third plenary session in July. To advance this vision, policymakers are promoting the ‘AI Plus’ (AI+) initiative, which echoes efforts dating back to 2015 to transform China’s economy and society through internet and information technologies. Part of the plan involves driving demand for large language models (LLMs) and other AI systems in traditional industries to improve efficiency.

 

 

 

 

Despite Beijing’s enthusiastic embrace of techno-solutionism, history shows that transforming an economy into a high-tech powerhouse is not an easy task. AI is no silver bullet, and machine intelligence alone is unlikely to solve the structural issues in China’s economy. In fact, Japan’s developmental model, which Xi implicitly referenced with his “head goose” analogy, failed to reach its full potential. Similarly, the success of the CCP’s technology-driven growth strategy is far from assured; investments in AI and other emerging technologies will need to stimulate productivity across other sectors. Previous efforts to promote "smart manufacturing" serve as a cautionary tale. Jeffrey Ding, a leading scholar on China’s technology and AI strategy, refers to this as a “diffusion deficit,” argues that China suffers from a “diffusion deficit,” noting that the country’s ability to adiot scientific and technological advancements lags behind its innovation capacity.

While it’s tempting to hype China’s growing innovation capabilities in AI, using metrics like the number of high-impact academic papers or computer science graduates, the reality is more complex. For example, it’s no surprise that Chinese government policy is focusing on talent and education, as the country still faces a shortage of qualified AI workers despite its rapidly expanding talent pool. Additionally, China struggles with a skills mismatch and youth unemployment remains high. Although it may seem counterintuitive for a country where 39% of the population will be over retirement age by 2050, the rapid pace of automation could have disruptive effects on China’s fragile labor market.

 

China faces a tradeoff between its AI and climate ambitions

 

AI could destabilize not only China’s labor market but also its energy systems and infrastructure. Despite Big Tech’s reluctance to address the issue—let alone disclose the power consumption of its data centers—AI, particularly compute-intensive large language models (LLMs), places enormous strain on natural resources and energy systems. The International Energy Agency (IEA) projects that Chinese data centers will account for nearly 6% of the nation’s total electricity demand by 2026. Additionally, generating electricity and cooling these data centers requires vast amounts of water. The Hong Kong-based China Water Risk estimates that total water usage by data centers in China could exceed 3 billion cubic meters by 2030—roughly equivalent to the entire annual residential water consumption of Singapore. China’s “war of a hundred AI models” could lead to wasteful competition for already scarce computing resources, threatening to derail the country’s green transition.

 

For China, reconciling its AI ambitions with its climate goals presents a monumental challenge. Beijing aims to peak CO2 emissions by 2030 and is transitioning to a two-pronged emissions reduction approach: rather than simply limiting energy consumption, China will control both carbon intensity per unit of GDP and total greenhouse gas emissions. Despite leading the world in renewable power generation, structural socioeconomic factors and grid bottlenecks mean that China still relies on coal for two-thirds of its energy mix. As computing infrastructure rapidly expands to meet the growing demand for LLMs, there is a risk that China’s energy systems won’t keep pace with the AI boom. The government is attempting to address this by relocating data centers and computing hubs closer to cleaner and cheaper energy sources, setting increasingly stringent energy intensity targets, and promoting better coordination of computing resources.

 

To be sure, AI could also present opportunities for China’s energy sector. The concept of a “smart energy brain” has gained traction as policymakers and state-affiliated researchers promote the synergistic development of computing power, AI, and the energy economy. One state-led project, the Tianshu-1 system, reportedly reduced energy consumption by over 15% by integrating AI and big data for tasks such as prediction, management, and maintenance. Chinese LLM developers are also seizing this opportunity, aiming to attract new customers and create models tailored to specific industry applications. For example, China Southern Power Grid has partnered with Baidu to develop a series of AI models for the power sector. However, the success of these initiatives is far from guaranteed.

 

 

 

US export controls on computing resources hamper China’s AI development

 

These domestic challenges are further complicated by external factors, particularly China’s reliance on American semiconductor technology for its AI development. An increasingly zero-sum “AI arms race” has become a key aspect of the strategic great power competition between China and the US. In October 2022, the Biden administration began restricting exports of advanced semiconductor technology to China, including cutting-edge graphics processing units (GPUs)—integrated circuits capable of performing high-speed mathematical calculations critical for machine learning tasks. The controls also extend to the tools, software, and expertise needed to produce advanced chips. Triggered by Beijing’s aggressive military-civil fusion strategy and authoritarian surveillance programs, these restrictions were tightened further in October 2023, with more expected to follow.

 

US export controls, which apply extra-territorially, place additional strain on China’s power generation by forcing Chinese firms to rely on larger quantities of older, less efficient chips for AI tasks. The CEO of DeepSeek, a Chinese LLM developer, admitted that indigenous LLMs require four times the computing resources of their US counterparts to achieve results that are still one generation behind. A recent study by scholars at Yale University estimated that if China had access to export-restricted chips for its data centers, the energy saved could equal the yearly power consumption of between 12,000 and 67,000 American households. Beyond hardware optimization, the study also found that a protectionist scenario makes algorithmic improvements less likely, potentially causing both American and Chinese Big Tech to waste the equivalent of the annual energy consumption of 1.8 million US households.

 

With consolidation in China’s generative AI landscape still far off, many players are competing for the limited computing resources available. Encouragingly, some academic and corporate labs are pursuing advanced research into brain-inspired intelligence. Neuromorphic approaches, which prioritize less energy-intensive models based on the brain’s structure, offer an alternative to the current trend of scaling up neural networks. This is a promising field internationally, and Chinese scientists are making notable progress. However, amid intense geopolitical competition to build bigger and better models, a broader reconsideration of China’s inefficient AI development strategy seems unlikely.

 

 

What China’s AI strategy means for Europe

 

Despite these challenges, China’s advancements make it the most formidable competitor in the AI race alongside the US. European policymakers, lawmakers, companies, and civil society can no longer afford to overlook China’s AI ecosystem, particularly as it progresses toward developing frontier AI systems. As Europe strives to degend its position in the geopolitics of technology, there are at least two key priorities to consider.

 

First, as highlighted by MERICS research, the European and Chinese AI ecosystems are more interconnected than often assumed, particularly through research collaborations. However, China’s state-driven approach, geopolitical objectives, and ambitions for dominance in this field call for a risk-based approach to partnerships with China-based companies, universities, and research organizations. US government policies and regulations—some of which apply extra-territorially, or may in the future—further complicate the delicate balance between national security, ethical technology development, and competitiveness. European governments have an important role to play in ensuring that Europe retains its agency in the AI landscape, not only by fostering AI innovation but also by protecting local talent and technologies from being siphoned off by American and Chinese Big Tech.

 

Second, Europe needs to articulate a clear vision for how it wants to engage with China on global AI governance. China’s government has implemented some of the world’s most ambitious AI regulations and is pursuing a proactive, two-pronged AI diplomacy. This approach positions China as a champion of the developing world while simultaneously engaging with the West on shared concerns about safety and potentially catastrophic risks. Bilateral talks with the United States on AI risks are ongoing, despite divergences. Additionally, China co-signed the Bletchley Declaration, the outcome of the AI Safety Summit hosted by the United Kingdom in the fall of 2023. Yet, with few exceptions, the EU has shown little interest in looking beyond political and value differences to better understand—and selectively engage with—China’s approach to regulating this powerful and transformative technology.