Longgang does not look like a policy laboratory. It is a manufacturing district of 413 million square metres of floor space and 30,000 industrial firms, the kind of place where the biggest employer makes LED panels and the second biggest makes supply-chain software.
On 28 September the provincial reform committee approved it as a pilot zone for artificial intelligence and robotics applications, the first in Guangdong built around deployment rather than research. On 15 October the district held its mobilisation meeting. By 28 October it had published a list of application scenarios and opened bidding on them.
The first market is health, not factories
The district’s own numbers explain the sequencing. Longgang has 4.13 million residents and about 25 million outpatient visits a year, roughly half of them at primary care level. That is a demand pool large enough to justify building products before anyone signs a national contract.
In May 2025 the district launched a health management service called Longgang Family Doctor across 12 public hospitals and 171 community health centres. Residents book a family doctor, read their own results and get an alert pushed to a relative when a reading moves outside range.
The hardware came from a local firm. Xinrui Technology built a health-management watch that the company says is the first domestic wearable to hold both a Class II medical device registration and a network access licence from the Ministry of Industry and Information Technology. It also sells a continuous glucose monitor that takes a reading every three minutes, several hundred data points a day, and the company argues that those readings are clinically usable rather than consumer-grade estimates.
This year the district paired Xinrui with Ant Group’s health assistant platform. The device collects, the model interprets, the platform issues a weekly report and books an appointment when a doctor is needed. The division of labour is the point: a hardware company does not have to become an AI company to sell into a hospital.
Firms post the problem, the district posts the money
The second mechanism attacks a failure mode that every industrial policy runs into. Technology firms have capability and no venue. Venues have pain and no vocabulary.
In September the city’s AI adoption roadshow stopped at Tianan Cloud Park in Bantian. More than 70 organisations attended. Three companies posted requirements on the spot: Absen Optoelectronics, Langhua Supply Chain and Longgang Data Group, eight problems in total, covering smart manufacturing, cross-border supply chains and data operations.
Absen’s request was the most specific. The company has led China’s LED display exports for 16 consecutive years and sells into 165 countries. On site, hundreds of modules have to be assembled by hand into a single wall, and the tolerance between adjacent modules is measured in hundredths of a millimetre. A small drift shows up as a visible seam.
On 28 October the district’s AI and robotics office put that requirement on the 2026 scenario list. Companies apply, experts review, results are published. A recognised project can receive up to RMB 4 million under current policy.
The direction of the paperwork has reversed. Firms used to apply to government for incentives. Now they write the requirement into a reform list and the government goes looking for a supplier.
What the base actually supports
None of this would matter in a district without suppliers. Longgang has 3,720 national high-tech enterprises and above-scale industrial output above RMB 1 trillion for two consecutive years. More than 800 companies sit somewhere in the AI supply chain, and the robot cluster is one of the densest in the country.
The two-step target is explicit. By 2028, AI should be embedded in six priority industries, smart terminal penetration should pass 90 per cent, a national AI open-source community should be operating, 1,000 benchmark scenarios should be live and total industry scale should exceed RMB 100 billion. By 2030 the district wants a full technical stack and a functioning ecosystem of suppliers.
Two caveats are worth keeping. Scenario lists are easy to publish and hard to fill, and a RMB 4 million subsidy is small against the cost of a production line retrofit. The test will be whether named firms with named tolerances get working machines, not whether the list gets longer.
Still, the sequencing is unusual. Most robot programmes start with a model and look for a use. This one started with a defect rate on an assembly floor and worked backwards.
Editor’s note: This is an adapted translation of the original Shenzhen News report. It has been trimmed and restructured for readability for an international business audience. The full original (in Chinese) is at https://www.sznews.com/news/content/2026-11/15/content_32214920.htm.