China
92 projects in 11 cities
The atlas documents 92 projects across 11 cities in China, the earliest dating to 2007: 13 with a documented success, 79 still running without a verdict. Every entry carries an evidence grade — the strongest here is peer-reviewed (6 entries) — and links its sources.
What the record shows
The Chinese record reads like one experiment run at national scale: let a single platform collect everything, then publish the results yourself. Alibaba's City Brain set the pattern in Hangzhou in 2016 — operator-reported gains, at what its own lesson calls a privacy price most democracies won't pay. The superlatives arrive pre-attributed: state media describe Yangshan Phase IV as the world's largest single-phase automated container terminal, and authorities call Tianjin's Section C the world's first zero-carbon smart terminal. Outside checks are rare and conspicuous: ITDP's independent 2011 evaluation credited Guangzhou's BRT corridor with travel times 29% better for bus riders and 20% for motorists; UNEP put Beijing's PM2.5 fall at about 35% between 2013 and 2017 — a whole programme's work, not the sensors' alone; and Shenzhen became the first city to fully electrify its buses — over 16,000 by 2017.
The instrument of choice is one login, one screen. Shanghai's Suishenban puts over 3,150 service items into a single app and reports more than 89 million registered individual users; its operations twin, One Network Unified Management, fuses cameras, sensors and agency records into a live digital twin — and scarcely a city in this record lacks its own version of one or the other. Nanjing's super-app shows how to read the numbers: 11.25 million cumulative registrations against 9.58 million residents — not fraud, the entry notes, but a decade of one-off sign-ups, the only figure that never falls. The sharpest edge is the league table: Beijing's 12345 hotline routes every complaint as a tracked work order and ranks districts monthly, and the peer-reviewed finding is double-edged — service delivery improves, but accountability strengthens upwards to superiors rather than to the public. The hardest limit met was public: Hangzhou's health code became the template for all 31 provincial-level administrations within 39 days, but when the same health commission tried to turn it into a permanent 0–100 lifestyle score, the backlash forced an immediate retreat.
No thread recurs here like the grammar of "driverless". Shanghai's Line 10 opened in 2010 with driverless-capable signalling and carried drivers for its first decade; by December 2021 the operator counted 167 km of GoA4 route across five lines and said that made Shanghai first in the world for driverless-metro scale — the same entry keeps the 2011 collision, 271 people injured after a power cut. Beijing's Yanfang line, described by state media and its operator as China's first fully automated metro built entirely from domestic subsystems, carried an attendant until 19 December 2019, going unattended only after 195 tests. Nanjing's Line 7 states the rule: the announcements say the trains 具备无须配置司机的能力 — have the capability to run without a driver — which is a statement about the design, not about who is aboard, and no accessible source names the day staff stopped riding. The counterexample: Shenzhen's lines 20 and 14 have run unattended since day one, the claim checked by TÜV Nord rather than taken on the operator's word.
The robotaxi record separates permits from services. Wuhan became the first Chinese city with fully driverless commercial rides in August 2022 — while its own transportation bureau counted 400 active vehicles in 2025 and called claims of 1,000-plus "rumors". Chengdu's ran for years as eight cars with a paid safety officer in every seat. Shenzhen wrote China's first dedicated autonomous-vehicle law in 2022 and issued a citywide fully driverless permit in October 2025 — and none of it insulated anyone when a cloud-side outage froze more than 100 Wuhan cars on 31 March 2026: no new permits nationwide from 29 April until an industry-wide safety review closed in late June, with expansion blocked even in Beijing's demonstration area, 38 million test kilometres in. The regulator now polices the vocabulary itself: after a fatal Xiaomi SU7 crash, the industry ministry barred carmakers in April 2025 from advertising "smart driving", "self-driving" or "autonomous driving", and made driver-assistance software updates subject to prior approval.
What sets this record furthest apart is what a city does with data. National doctrine since December 2022 — the "Data Twenty Measures" — splits data property into three separable rights and tells governments to supply public data "in the form of models and verification services", with original data remaining in domain, usable but not visible. Beijing's financial data zone is that sentence built: approved banks compute on municipal data they never receive, and the operator sells scores and verification services — never datasets — and reports over 400 million service calls. Shanghai's Data Exchange is the market half: over 2,700 data products listed by mid-2024 and RMB 1.16 billion traded in 2023 on the city's figures — while analysts, the entry notes, still question whether the venues have real liquidity. The portals themselves open with conditions: Shanghai's publishes in tiers from unconditional through application-only to closed; Guangzhou's requires attribution, forbids resale and asks users to register their intended use; Hangzhou's sits in the top tier of Fudan University's independent open-data index and still keeps downloads behind a registration wall — a narrower kind of open.
Where a rule already binds, the digital layer usually came second — and the record grades the rules, not the software. Beijing has rationed the right to buy a car by ballot since 2011: a peer-reviewed study exploiting the lottery's own randomisation found the restriction cut the city's car stock by 14%. Guangzhou's lottery-plus-auction shows the design margin — its published evaluations point to substitution rather than restraint. Shanghai's waste-sorting mandate grew its digital layer around the rule rather than with it — AI cameras grading bin-room purity — while the points scheme meant to reward good sorting largely stalled. The strongest entries connect a reading to a consequence: Chongqing's bridge sensors include weigh-in-motion units certified precisely enough for police to cite overweight trucks without a roadside stop, while Guangzhou's 6,860-station drainage network still forecasts floods for a human crew to act on. If these entries agree on one lesson, it is to read the operator's grammar: a capability is not an operation, a registration is not a user, a listing is not a market — and a reading becomes a record on the day someone outside the operator is allowed to check it.
Sources beyond the atlas
Facts above that do not come from a linked entry rest on these:
- CPC Central Committee & State Council — Opinions on Building Basic Systems for Data ("Data Twenty Measures"), December 2022 — the three separated data-property rights, public data supplied "in the form of models and verification services" with original data remaining in domain, and the multi-tier data-trading-venue system
- TechCrunch, 17 April 2025 — reporting an MIIT directive first seen by Reuters — carmakers in China barred from using "autonomous driving", "self-driving" and "smart driving" in marketing, with driver-assistance software updates made subject to government approval, after a fatal Xiaomi SU7 crash
The shape of China
Computed from the 92 entries below — a profile, not a score.
National programs in this record
Filed under the city where the system demonstrably runs — the atlas's scope rule for national-scale programs.
Cities
All 92 projects
✔ Every published entry is checked against the public sources it cites before publication.
Show 80 more projects in ChinaShow fewer
Deeper reads
Where China appears in the atlas's own analysis.
📎 Cite this country record
1001 Smart Cities (2026). “Smart-city projects in China.” The Smart City Atlas. https://1001smartcities.org/countries/china/ (last verified 2026-08-28). Data: CC BY 4.0.
The underlying data is free to reuse with attribution — see the open data page.
