AI Rewrites the Logic of Cities as Tsinghua Management Review Calls for a Governance Shift

Huxiu published an essay in October 2026 from Tsinghua Management Review, a publication affiliated with Tsinghua University’s School of Economics and Management. The authors, including Lin Lan, frame AI’s impact on cities as more than “adding another intelligent shell to existing systems.” In their view, the temporal, spatial, and organizational logics of urban operations themselves are gradually changing.

Conventional urban management has followed a sequence of build first, govern second, and correct afterward. The essay argues that this is likely to shift toward a model of sensing while operating, adjusting in real time, and correcting continuously. The Huxiu article describes this change not as a momentary event but as a structural process that unfolds continuously.

What truly needs to be debated is no longer “whether cities use AI,” but how cities should organize and constrain AI and reshape themselves through it.

The focus is not on whether to adopt AI but on design and control. Below, we summarize the arguments in Tsinghua Management Review© and consider their implications for practitioners involved in urban systems.

Cities: From Physical Containers to

Learning Systems

Whereas industrialization formed cities centered on roads, factory buildings, pipe networks, and zoning, the change in the AI era lies in cities beginning to acquire a “cognitive layer.” The Huxiu article explains that sensors, digital twins, large models, knowledge graphs, and predictive algorithms convert flows of people, goods, energy, and information into analyzable, callable data resources. This changes how urban problems are identified, interpreted, and handled.

On the governance side, a structure emerges in which government uses models to allocate resources, provide early warning of risks, and optimize operations. Companies, developers, and residents can also participate in invoking urban-development capabilities and reconfiguring tasks through open interfaces, platform tools, and feedback mechanisms. As a result, the essay argues, cities shift profoundly from “fixed upon completion” to “continuously updated in operation.”

What matters here is that AI’s effects are not confined to operational efficiency. Because the three logics of time, space, and organization are reconfigured simultaneously, the premises of procurement and organizational design also change. Urban information systems must shift their evaluation criteria from comprehensiveness of functions to updatability.

Lifecycles Shift from Linear to Iterative

For a long time, the urban lifecycle followed a linear path: plan—build—maintain—renew. Once infrastructure was built, it entered a relatively static operating phase, and actual adjustments began only after problems such as congestion, aging, pollution, heat islands, and safety accidents had accumulated. This structure pushed problem discovery and repair to the end of the chain.

AI pushes this delayed form of management into a new cycle of learning and correcting while operating. With digital twins, real-time sensing, and model-based simulation support, cities can accumulate feedback during operations, revise parameters, and update rules. Maintenance and renewal that once lagged can be moved forward into daily operations.

The objects of maintenance also change. In addition to the physical lifespans of bridges, subways, and drainage systems, it becomes necessary to monitor model accuracy degradation, data distortion, interface aging, and coordination failures between departments. A system that handles physical and cognitive maintenance simultaneously is required.

Algorithms Determine the Half-Life

A city’s “half-life” no longer depends solely on the pace of physical depreciation. It is also shaped by the cycle of iterative algorithm updates, data quality, and governance feedback mechanisms. The essay compares the city not to a machine that runs stably for a long time once built, but to a complex operating system that requires continuous patching.

Energy also comes to the fore. Training and inference for large models and operating data centers require vast computing resources and electricity, and cities must redesign the relationship among “computing resources—power—heat recovery.” Securing dedicated green power, recovering waste heat, and optimizing loads are not merely technical issues for the energy sector; they are opportunities that shape the transformation of the urban lifecycle.

Which cities can first embed computing infrastructure into systems of green energy and district heating will determine whether AI expansion becomes a new burden of high energy consumption or an increment to low-carbon transition. Data center siting is shifting from a communications-network design issue to an urban-planning issue.

Reorganizing Economy,

Governance, Society, Culture, and Space

The essay organizes urban reconstruction across five layers. The economy shifts from accumulating factors to being model-driven. Governance is compelled to convert model legitimacy into institutional legitimacy. Society reorganizes from a system of positions to a system of tasks, while culture shifts its center of gravity from the abundance of generation to the maintenance of shared reality. Space changes from competition over location to competition over network position within a composite skeleton of digital and energy infrastructure.

The variables used to measure urban competitiveness also change. The emphasis shifts from visible factors—land-development intensity, scale of capital investment, level of population concentration, total infrastructure—to invisible capabilities such as corpora, models, evaluation, interfaces, rules, scheduling of computing resources, and coordination capacity. What matters is not the total volume of equipment but whether capabilities can be continuously updated.

This framework has the effect of raising the smart-city debate from the level of adding functions to the level of institutional design. The number of sensors installed and the sophistication of dashboards become insufficient indicators. Practitioners will need to judge whether systems are structured so models can be swapped and whether external parties can verify the basis of evaluations.

Shanghai’s Strength Lies in Rule-Making

The fact that Shanghai is cited as a comprehensive case raises many practical issues. What the essay praises is not innovation in application scenarios but the rule-making dimension. It builds industry corpora, evaluation benchmarks, and computing-resource scheduling as shared public infrastructure, and for high-impact areas it provides algorithmic impact assessments, channels for objections, and mechanisms for suspension and removal.

The argument is that shifting the emphasis from “mastering AI” to “being able to govern AI” is the condition for converting an early advantage into a long-term one. The source of competitiveness is not the quality of individual applications but institutional design that makes it possible to swap or stop models.

Shared evaluation benchmarks and computing resources also change the dynamics of procurement. As municipalities move from selecting vendors individually to comparing multiple models on a common platform and switching when necessary, vendors are increasingly required to ensure explainability.

How to Contain Four Bottom-Line Risks

The essay identifies four risks: constraints on computing resources and energy, a crisis of trust in data, dependence on platforms, and systemic vulnerability. All tend to become visible in the operational phase after adoption and are difficult to capture through ex ante performance evaluation alone.

The uneven distribution of computing power directly translates into disparities between cities. Models that cannot explain the provenance and quality of their data are hard to use as a basis for administrative decisions. Dependence on a specific vendor erodes procurement autonomy, and in tightly coupled systems a local failure can spread widely. The four risks are not independent; they amplify one another.

What the essay emphasizes here is not using risk as a criterion for whether to adopt AI but building it into the design conditions for operations. Systems built on the premise that they can be stopped are better able to limit damage during failures. Conversely, systems that cannot be stopped carry large costs for correcting errors.

Design Principles:

Exit, Degradation, and Substitution

The essay sets out “exit, degradation, and substitution” as design principles for the four risks. Risk management must be embedded across the entire urban-system lifecycle, and it is necessary to ensure that an entity exists to explain, take responsibility, and correct errors after they occur. At the core of the principles is the need to avoid a vacuum of responsibility.

The use of the terms co-governance and stewardship should not be overlooked. It envisions a third form that is neither control solely by government nor full delegation to vendors. It is a framework in which citizens, businesses, and government share roles and assume long-term management responsibility.

For Japanese municipalities, this is an issue directly tied to contract and operational design. Whether contracts can specify obligations to update models, correction procedures when performance degrades, and data migration upon withdrawal will determine the success or failure of implementation. Before selecting technology, it is necessary to prepare templates for such clauses.

This article is based on Huxiu (All Rights Reserved). It relies on fair quotation under Article 32 of the Copyright Act of Japan.

Editorial Opinion

In the short term, we expect stronger moves to add items such as algorithm registries and impact assessments to municipal procurement specifications. The evaluation axis for urban infrastructure will shift from comprehensiveness of functions to explainability, and vendors will be pressed to develop evaluation benchmarks. The design of demonstration experiments itself may be called into question.

In the long term, constraints on computing resources and electricity will become preconditions for urban planning. Data center siting and waste-heat use will rise on the urban policy agenda, and the maintenance costs of model updates will be recognized as fixed costs for municipalities. The practical focus will be how to write update and suspension clauses into procurement contracts.

As an editorial team, we want to ask who holds the authority to suspend or remove a system. When operations are delegated to the private sector, responsibility for model accuracy degradation tends to become ambiguous. In Japan, institutional design at the demonstration stage is thin, and this needs to be examined.

References

• “ 当AI进入城市：城市发展的范式转向与治理重构 “, by 清华管理评论© — 虎嗅网, 2026-10-18T22:16:54.000Z (ARR)
• Source URL: https://www.huxiu.com/article/4898451.html?f=rss