China's Science Funding Call Signals a Hard Push on AI-for-Discovery
China's National Natural Science Foundation (NSFC) is doubling down on AI as a core engine for scientific discovery. The latest annual call shows a clear theme: blend AI with real-world experimentation to push scientific frontiers-and inch closer to systems with human-like problem-solving.
This isn't just about better tools. Some of China's leading AI thinkers argue that doing science is the ultimate benchmark for AGI, with research labs building the models and compute to make it real.
AGI Through the Lens of Science
Shanghai AI Lab's director Zhou Bowen has been blunt: if AGI is coming, it will be built on scientific ability. In this view, AI must hypothesize, test, and explain-not just predict.
Policy aligns with that logic. In proposals shaping the 15th Five-Year Plan, Beijing called for AI to "lead a paradigm shift in scientific research," setting the tone for how grants and labs will operate in the next cycle.
What the NSFC Is Funding Now
- Generative AI for designing complex fusion reactors-aimed at automating high-stakes engineering loops.
- A design platform for biological components that can withstand extreme environments, targeting bottlenecks in biomanufacturing.
- A sub-program titled "explainable and general-purpose next-generation AI methods" (可解释、可通用的下一代人工智能), hinting at tight coupling with national next-gen AI initiatives.
About half of the listed project categories reflect this AI-for-science push. It's a systematic prioritization, not a one-off experiment.
Small Grants, Big System
Per-project funding looks modest-roughly EUR 240,000 to 600,000-but new awards roll out annually. The long game matters more than a single grant size.
Behind the scenes, frontier labs are building large models and compute clusters to fuel this strategy. Chinese teams are applying AI to control fusion plasmas, map the universe, and infer physics principles with minimal prior input.
Where This Fits: 2030 Megaprojects and Beyond
The NSFC's "Major Research Program" (重大研究规划) is one of several vehicles backing strategic science. It likely interlocks with China's 2030 S&T megaprojects, announced in 2017, even if the implementation details remain tight-lipped.
The throughline: AI isn't a sidecar-it's the engine for priority fields where faster discovery cycles translate into national advantage.
Signals Policymakers and Research Leaders Should Note
- Bench-to-model integration: Expect funding to favor projects that combine simulation, lab automation, and explainability.
- Compute as infrastructure: Model training capacity will determine which labs move from prototypes to platforms.
- Explainability isn't optional: Programs explicitly call for interpretable, general methods-not black boxes.
- Fusion and bio as templates: High-complexity domains are early proving grounds for AI-driven design loops.
Practical Moves for Your Organization
- For research teams: Pitch projects that close the loop-AI model + simulation + automated experiment + theory extraction.
- For government funders: Pair grants with shared compute and data standards to avoid siloed efforts.
- For lab directors: Invest in tools for explainability and scientific reasoning, not just bigger models.
- For industry partners: Watch fusion, materials, and biomanufacturing as early adopters of generative design.
What to Watch Next
- More "AI-as-scientist" benchmarks that test hypothesis generation, experiment planning, and causal reasoning.
- Deeper links between grant topics and national megaproject milestones leading up to 2030.
- Growth in shared platforms-data, simulators, and lab automation-that standardize AI-for-science workflows.
If you track AI policy or run a research portfolio, this call is a clear cue: fund and build systems that turn models into measurable scientific progress. The institutions that nail the full stack-data, compute, methods, and lab execution-will set the pace.
NSFC (official site) | Shanghai AI Laboratory
Your membership also unlocks:
