China faces a fundamental constraint in its pursuit of artificial intelligence-driven scientific breakthroughs: it cannot produce the sophisticated equipment necessary to generate the high-quality experimental data upon which advanced AI models depend. The challenge became starkly apparent last week when Weinan E, a Peking University mathematician and Chinese Academy of Sciences member who conceptualised the "AI for Science" framework in 2018, warned that without domestically manufactured precision instruments, researchers struggle to obtain the raw material essential for training powerful scientific models. His colourful metaphor—that Chinese science operates "like cooking without rice"—encapsulates a structural vulnerability in Beijing's strategy to leapfrog Western scientific dominance through algorithmic innovation.
The scope of China's import dependence is substantial and across multiple critical domains. In 2024 alone, China imported nearly US$17 billion in scientific equipment, with more than three-quarters of major research instruments deployed domestically sourced from abroad, according to analysis by Puhua Policy, a Beijing-based consulting firm. The gaps are particularly acute in instruments fundamental to contemporary research. LeadLeo, another consultancy, found that China relies on imports for 83 per cent of mass spectrometers and chromatographs—devices essential for identifying molecular composition and separating chemical compounds for analysis—and 75 per cent of spectrometers used to study material properties through light-based analysis. Furthermore, China is almost entirely dependent on overseas suppliers for optical instruments and biological tissue analysis equipment, creating vulnerabilities across interconnected research domains.
This dependency carries practical consequences that extend beyond mere technical limitation. The reliance on foreign suppliers translates into elevated equipment costs, protracted maintenance cycles, and delayed after-sales support that collectively undermine research efficiency. Supply-chain fragility compounds these concerns, particularly given geopolitical tensions. Should foreign suppliers restrict access or face disruption, China's scientific infrastructure could face simultaneous equipment failures and repair backlogs across multiple facilities. For Southeast Asian research institutions seeking to collaborate with or learn from Chinese scientific advances, this limitation indirectly affects the region's capacity to participate in cutting-edge research partnerships.
The United States has weaponised this structural vulnerability through systematic export controls. During Donald Trump's first presidency, by December 2020, over 42 per cent of China-related entries on the US export control list involved scientific and precision equipment. These restrictions have intensified under Trump's second term, driven by American concerns that advanced instrumentation could inadvertently support Beijing's military modernisation efforts and accelerate the development of AI-assisted weapons design. In January, the US Department of Commerce announced fresh export restrictions specifically targeting high-parameter flow cytometers and certain mass spectrometry equipment, explicitly citing their capacity to "generate high-quality, high-content biological data, including that which is suitable for use to facilitate the development of AI and biological design tools."
These export restrictions represent a deliberate strategy to constrain China's ability to acquire the data infrastructure necessary for competitive AI development. By limiting access to precision instruments, Washington effectively restricts the volume and quality of experimental data available to Chinese researchers, making it substantially more difficult to develop and validate advanced scientific models. The strategy operates at the foundation level—blocking inputs to the research pipeline rather than targeting outputs. This approach reflects sophisticated understanding of how contemporary AI development depends not on algorithmic innovation alone, but on access to vast quantities of high-fidelity experimental data that only precision instruments can generate.
Parallel to equipment constraints, Weinan E identified a second critical weakness in China's AI-for-Science architecture: significant gaps in foundation models compared to international counterparts. He cautioned that adding scientific capabilities to existing open-source models represents a "false premise," arguing instead that solving genuinely complex scientific problems demands fundamentally stronger underlying models rather than post-training modifications applied to weaker base systems. This observation points to a deeper strategic divergence between American and Chinese approaches to AI integration with science. The United States concentrates resources on strengthening general-purpose foundation models and integrating them with automated research infrastructure, whereas China has pursued a more application-driven methodology.
China's approach involves constructing integrated scientific AI infrastructure that consolidates data, software, computing resources, and automated equipment, then applies these capabilities to specific research fields. While this bottom-up, application-focused strategy has merits in generating near-term practical results, it may sacrifice the versatility and transferability of more general-purpose systems. Weinan E argued this reflects a structural mismatch with how advanced AI systems develop—that powerful models emerge from broad foundational capabilities rather than narrow specialisation. For Malaysian and Southeast Asian institutions evaluating technology partnerships and research collaborations, this distinction matters considerably, as it suggests Chinese AI capabilities may excel in specific domains while lagging in broader scientific applications requiring model flexibility.
To address these interconnected vulnerabilities, Weinan E proposed a fundamental restructuring of China's research system for the AI era. He identified three necessary "breaks." First, scientific boundaries between disciplines must dissolve to enable cross-field research, recognising that transformative discoveries often emerge at disciplinary intersections. Second, the traditional divide between theoretical research and experimental work requires dissolution, as AI-driven science depends on continuous feedback between modelling and data collection. Third, barriers between academic institutions and industry must lower, as advanced research infrastructure increasingly resides in commercial settings and successful implementation demands seamless knowledge flows.
Beyond structural reorganisation, Weinan E advocated for overhauling traditional research evaluation systems that privilege academic publications above all other contributions. Contemporary science increasingly depends on developing shared data repositories, open-source software tools, and automated research infrastructure that enable future discoveries but rarely garner recognition in conventional publication metrics. Reorienting evaluation systems to value these foundational contributions would align incentive structures with the requirements of AI-driven research, potentially accelerating both domestic capability development and capacity for international collaboration. For Southeast Asian research institutions and policymakers, this insight carries relevance: the region's scientific competitiveness increasingly depends not merely on individual researcher brilliance but on institutional capacity to build and maintain shared research infrastructure.
The convergence of equipment dependence, export restrictions, and foundation model gaps creates a strategic trilemma for Chinese scientific ambitions. China cannot simultaneously achieve research self-sufficiency, maintain rapid progress in AI-driven science, and operate within current international trade frameworks. Some combination of domestic instrument manufacturing acceleration, negotiated trade agreements protecting scientific access, or fundamental repositioning of research strategy appears inevitable. These choices will reverberate across Asia's scientific landscape, influencing which research hubs dominate emerging fields and where collaborative opportunities concentrate. For Malaysian institutions and Southeast Asian policymakers, China's struggles and responses offer cautionary lessons about technology dependence and strategic planning implications.
