The next big competition in artificial intelligence is over science. The question is shifting from who has the largest model to who can use AI to shorten the discovery cycle: forming a hypothesis, running simulations, designing an experiment, operating equipment, collecting results and learning from them.
The United States made that shift explicit. The Department of Energy’s Genesis Mission aims to bring together 17 national laboratories into an integrated discovery platform to double the productivity and impact of US research and innovation within a decade. This year, the programme announced a partnership with Japan, reinforcing the idea that AI for science is becoming infrastructure and a network, not simply another model.
China is moving in the same direction but increasingly through cities. Beijing’s 2025 plan calls for common scientific-intelligence infrastructure, at least 10 high-quality scientific databases, more than 10 million users and at least eight benchmark applications by 2027. Shanghai’s “Hundred Teams, Hundred Projects” initiative offers support for up to 70 per cent of approved investment, capped at 50 million yuan (US$6 million). The race to harness AI for science is no longer merely national.
This creates both an opportunity and a trap for Shenzhen.

The opportunity is obvious: AI could become a new source of scientific productivity and industrial growth. The trap is copying what other cities are doing. Shenzhen should not try to become Beijing with more hardware, or Shanghai with a different set of industry labels; it should build around its own comparative advantage.

The first advantage is the structure of its innovation economy. Shenzhen invested 245.31 billion yuan in research and development in 2024, the highest R&D intensity among Chinese cities. Strikingly, companies accounted for over 93 per cent of that spending. Those figures describe a city where research, engineering and commercial deployment sit unusually close together. This matters in the AI-for-science drive because the hard part isn’t only inventing an algorithm, but turning it into a reliable scientific tool.