Is China a threat, or just out-engineering everyone?

The eminent scientist and sinologist Joseph Needham famously formulated two questions on Chinese science and technology (S&T) in his 1969 book The Great Titration: Science and Society in East and West: 1) Why did modern science not develop in Chinese (or Indian) civilisation but only in Europe? 2) Why was Chinese civilisation much more efficient than the West in applying human natural knowledge to practical human needs, up until the 15th century?

For the first question, Needham noted that the Confucian dogma of the strong central bureaucratic state had constrained the growth of commerce and industry in early modern China. Moreover, he argued that Chinese science was based on “the philosophy of organism”, where the universe was seen as a harmonious, interconnected whole without an external creator. In contrast, European mechanistic natural philosophy saw the universe as a machine created by a divine lawgiver and thus subject to discoverable mathematical laws and experimentation — core aspects of modern science. 

Inverting Needham’s questions

But this paper is not concerned with Needham’s questions in their original sense. It might be more relevant today to discuss an inverted version of the questions, namely 1) Why is China developing new contributions to global science and innovation? 2) Why is there an alternative dynamism in Chinese innovation? 

For decades, influential Western observers have lambasted China’s economy as technologically stagnant, lacking creativity in research and innovation, dependent on Western knowledge and technology, and supported by unfair subsidies from the government. While more analyses today acknowledge a new dynamism in the development of S&T in China, they are often portrayed in militaristic headlines, such as a “US-China Tech War” or the “Thucydides Trap” where a rising power threatens to displace an established great power.

That said, some pieces offer a more insightful reading, such as Dan Wang’s Breakneck, which argues that China is an engineering state, which brings a sledgehammer to problems both physical and social, in contrast with America’s lawyerly society, which brings a gavel to block almost everything, good and bad. 

China, a science and innovation juggernaut 

Since the turn of the century, the Chinese leadership has supported a surge in research and development (R&D) expenditures, which grew by over 23 times between 2000 and 2022. Moreover, China’s gross expenditures on R&D (GERD) in 2023 reached 96% of US GERD, up from 72% in 2013. Central government spending on S&T is set to increase by 10% in 2026, even as other areas face budget cuts. The 2026 Government Work Report set a target for nationwide R&D spending to increase by an average of at least 7% annually during the 15th Five-Year Plan (2026-2030). 

China’s share of global R&D expenditures has thus surged from 4% in 2000 to over 27% in 2025. Simultaneously, China’s R&D intensity as a percentage of GDP increased from 0.89% in 2000 to 2.8% in 2025, exceeding the average figure for OECD. According to a recent report, China is likely to eclipse the US as the world’s largest funder of public science by 2028.

The business sector has consistently been the main driver of R&D in China, accounting for a majority of total spending. Its share of GERD has increased from around 70% in the late 1990s to 76% in 2023. As Trelysa Long’s ITIF report shows, US firms expanded R&D investments by 150% from 2014 to 2024, while Chinese firms increased theirs by an extraordinary 537%. In contrast, firms in the rest of the world increased their R&D investments by just 32%.

Furthermore, the effective value of these Chinese R&D investments is further enhanced by the availability and relatively low cost of human resources. China’s researcher workforce is the largest in the world in terms of the number of full-time equivalent (FTE) research positions, having overtaken the US since 2009 and the EU in 2018. 

In 2018, China overtook the US in terms of the total number of science publications, according to statistics compiled by the US National Science Foundation (NSF). China has consistently increased its share of the 1% top scientific publications since 2005 and outpaced European countries (including Great Britain) in 2019.

Moreover, as Caroline S. Wagner has recently demonstrated, Chinese publications in the world’s elite natural science journals — including Nature, Science, Cell and Physical Review Letters — in 2024 reached a total of 37,273 articles, compared to America’s 31,930. In addition, China’s share in the world’s top 1% of most highly cited papers for 2024 had already achieved rough parity with the US, indicating that the quality of a significant number of Chinese publications was on a par with those from the US.

The field of artificial intelligence (AI) reveals much of the nature of innovative performance and the rivalry between the West and the East. Situating US and China in terms of scientific output and invention patents related to AI provides a general picture of how the two countries have developed their competences in AI. Besides, it is interesting to observe how policy makers and a range of actors in China and the US have made strategic priorities for the sector.

AI in China: growing R&D, patents

The development of AI in China took off during the 1990s, when the digitalisation of the Chinese economy underscored the need to develop AI. The competition between the leading digital platforms, known as BAT (Baidu, Alibaba, Tencent), created rapid development of AI for search engines in the first decade after 2000. Not only that, several companies emerged that developed AI for special services, e.g. Sensetime for facial recognition and iFlyTek for automatic translation. 

In 2017 the Chinese government issued a strategy for AI called “A New Generation Development Plan for Artificial Intelligence”. This strategy envisages three phases of development up to 2030 in which Chinese theories, technologies and applications of AI will become globally leading and China’s AI industry will be internationally competitive.

Chinese scientific publications in the field of AI have also grown significantly since then, reaching 24,000 publications by 2024. Indeed, Chinese output by volume in 2024 is equal to the sum of the EU-27, UK and US during the same year combined. Simultaneously, the quality of AI publications from China has increased. While Chinese publications received only a few citations in 2000, in 2024 the proportion of citations to Chinese research increased to just over 40% of global citations. The US and EU-27, by comparison, each received just around 10% of global citations in 2024, with the UK at a 2% level.

A similar picture emerges when the extent of patenting in AI is examined. From 2010 to 2023, the US patent and trademark office and the Chinese intellectual property administration granted 659,504 and 651,629 AI patents, respectively. Both countries have experienced rapid growth in AI patenting, with a particularly sharp acceleration in recent years. Thus, China has overtaken the US in the total annual number of AI patents granted since 2020.

China’s proactive policies make the difference

The government’s policy on AI has also evolved since the first AI strategy was announced in 2017. On 27 August 2025, China’s State Council issued a new action plan called AI+, which calls for more concrete ambitions for the country’s national AI strategy in the coming years. The aim is to achieve 70% AI adoption in key sectors by 2027 and 90% by 2030, with a further vision of building a fully AI-powered economy by 2035.

Chinese AI developers still struggle with access to the most advanced semiconductors for AI due to US export restrictions. This means they are lagging behind American tech giants in terms of facilities for compute. But on the other hand, China has built a more robust telecommunications and electric power grid and is expanding data centres in the interior provinces.

The Chinese leadership has chosen to see AI as a general-purpose technology that will be able to transform the Chinese economy, similar to how the steam engines transformed Europe during the age of industrialisation. China has been building an automation and robotics industry for decades, and the 15th Five-Year Plan declared a new focus on “embodied-AI”, which stands for the complete integration of virtual AI with hardware.

Moreover, the emphasis on open-source AI models in China allows firms to create systems for specific sectors, particularly designed to increase efficiency and effectiveness in the performance of producers. This emphasis has also allowed Chinese models to be exported as a key input for the development of appropriate AI in the Global South.

AI in the US: large investments to maintain lead

America occupies a position at the forefront of AI innovation, with strong competencies in both hardware and software. The evolution of AI technology was initially shaped by academic excellence, with the publication of AI-related scientific research results leading the world. As a result, American companies such as Nvidia and AMD have become leaders in the manufacture of advanced chips for training and utilising AI. Moreover, companies like OpenAI, Anthropic and leading tech giants have developed the most advanced large language models. 

Large investments in data centres with advanced AI chips help to secure American firms a leading position in future AI models and infrastructure. Underlying this approach are the competitive advantages in hardware, but American firms have also been at the forefront of advanced software for coding AI models. For example, Nvidia created the powerful parallel computing platform CUDA (Compute Unified Device Architecture) which is able to significantly accelerate tasks in AI. Another important AI related software is Code Llama, a state-of-the-art large language model for coding, which was developed by Meta AI.

The focus on large language models (LLMs) and generative AI (GenAI), and ultimately Artificial General Intelligence (AGI), has shaped the main applications of AI. For instance, generative models have been applied to drug discovery, and vast repositories of medical data are used to train diagnostic systems. LLMs have also been used in finance industries for customer service automation, and in entertainment industries for music composition, script development, and image or video generation. 

Low returns on AI and inefficient data centres

In spite of the initial enthusiasm that GenAI introduced in these sectors, however, a 2025 report from the Massachusetts Institute of Technology revealed that 95% of the US corporate world reported getting zero return on their AI projects. Only 5% of integrated AI pilot projects were extracting millions of dollars in value, while the vast majority remained stuck with no measurable impact for the organisation. Many enterprises had been eagerly piloting GenAI tools, but very few reached full deployment of AI, mostly due to integration complexity and lack of fit with existing workflows.

The dependence on large data processing facilities has further accentuated a problem of access to plentiful energy. Upgrading the US’s aged electricity grid — a fragmented, hybrid system primarily owned by private, investor-owned utilities — will need time and substantial investment. The plans to establish large data centres in various localities have already raised public concern about the feasibility of continued local supply. Similar issues have been raised about the impact of data centre cooling facilities on local water supply.

Donald Trump declared in 2025: “America is the country that started the AI race. And as President of the United States, I’m here today to declare that America is going to win it.” The Trump administration issued the four-page memorandum A National Policy Framework for Artificial Intelligence in March 2026, allegedly to protect children, safeguard American communities and enable innovation to ensure American AI dominance. However, this policy statement has been criticised as inadequate, both to promote the AI industries and to regulate the applications of AI.

This brief review of the comparative development of innovation and AI development in China and the US cannot offer a fail-safe bet on which place will “win”, but instead points to divergent strategies and performance. American advantages in foundational hardware and models provide the country with a leading position. However, in terms of the implementation of AI in the economy, the American experience may face new challenges. In contrast, China’s application of AI in automation capitalises on the extensive and advanced digital infrastructure and manufacturing background that the country has constructed over several decades. 

What does inverting Needham’s questions tell us?

When Joseph Needham set out on his quest to discover and document S&T in China, he was definitely not concerned with something like the “Europe-China tech war”, but with showing that the Chinese people had a long tradition of discoveries that had also helped technology development in Europe. The fact that European modern S&T helped the British defeat the Chinese empire in the Opium War of 1840 was, in his view, regrettable, not a triumph. Why do we need to be constantly told that China’s emergence as a major contributor to global S&T is a “threat” — a design to wage tech war against the West? 

The essence of my argument here is that inverting the Needham questions re-focuses the debate on China and the West to explain the new divergence —  exploring alternative development paths. What are the factors fuelling scientific achievements and innovative capabilities in contemporary Chinese civilisation, and do these factors provide an alternative global model for innovative dynamism?

Related: Cheap, fast and everywhere: Why China leads AI adoption | Rewriting the rules: Huawei’s new gamble to break the US’s chip blockade