As urban digitalization has progressed to this day, we seem to have fallen into a subtle dilemma.
On one hand, there is the surge in technology. From the explosion of large models like DeepSeek to various AI agents such as OpenClaw, AI (Artificial Intelligence) is penetrating every aspect of urban governance, public services, and industrial upgrading at an unprecedented speed.
On the other hand, the reality is harsh. The fragmented network systems of various local government departments, the bandwidth explosion caused by massive video access, and the extreme demand for millisecond-level latency from supercomputing centers... Traditional networks are becoming the biggest bottleneck for urban digitalization to be both fast and stable.
Meanwhile, national policies have set higher requirements for urban digitalization. The Guiding Opinions on Deepening the Development of Smart Cities and Promoting City-Wide Digital Transformation released in 2024, and the Action Plan for Deepening the Development of Smart Cities and Promoting City-Wide Digital Transformation in 2025, both clearly state: strengthen the construction of AI public service capabilities, and encourage mega and super-large cities to orderly promote AI-empowered city-wide transformation.
The national direction is clear, but when implemented in each city, how to take this step both fast and steadily remains an unsolved problem.
It is precisely against this backdrop that Huawei proposed the concept of "One Intelligent Network" at the 9th Digital China Summit. It attempts to answer a key question: when the training of urban large models requires the efficient flow of massive data, when data element trading demands end-to-end trusted circulation, and when cross-industry collaborative governance calls for deterministic network guarantees, what kind of network is needed to support all this? How can the network evolve from a pipeline to a foundation, enabling city-wide intelligence to truly land?
In early 2026, an AI application named "Lobster" suddenly went viral, allowing ordinary people to experience the magic of AI agents. Meanwhile, from the DeepSeek large model to various industry-specific AI agents, AI is penetrating every corner of urban governance, public services, and industrial upgrading at a speed beyond expectations.
However, behind the glamorous applications, an awkward reality is emerging: the network infrastructure in many cities is currently still unable to support the rapid surge in the AI era.
A frequently overlooked issue is: no matter how strong the computing power is, if data cannot be transmitted efficiently, its value will be greatly discounted.
In the past two years, a boom in the construction of intelligent computing centers has swept across various regions, with investments often reaching billions and computing power scales of hundreds or even thousands of Petaflops (P). However, within many cities, there are no high-speed channels between the data centers of different government departments, and the collaborative scheduling of computing resources between cities is even more difficult.
Government affairs, transportation, healthcare, education... Each department has its own business system and its own network. These networks were built by department in the past, with inconsistent standards and no interoperability. Data is locked in isolated islands, making it difficult to generate value.
As urban governance moves towards refinement, massive IoT (Internet of Things) terminals need to transmit data back in real time. Taking video as an example, urban governance in a medium-sized city requires the access of hundreds of thousands of video signals, resulting in a nearly tenfold increase in bandwidth demand. Many cities have built supercomputing centers and intelligent computing centers, but these "computing power hearts" lack high-speed channels between them. When scheduling computing power across regions, the latency reaches tens of milliseconds, making it difficult to meet the demands in a timely manner.
Security issues are equally noteworthy. In recent years, security incidents of government networks being attacked by hackers have occurred frequently worldwide. Hacker attack methods are becoming increasingly diverse, and boundary protection is no longer sufficient. Government networks carry a large amount of sensitive data, and the consequences of a leak are unimaginable.
A city is an organic "life form." Isolated AI cannot achieve true intelligence, and the network, as the most fundamental foundation, is in urgent need of a new upgrade.
To break this dilemma, we must start with network infrastructure. But the question arises: what kind of network does urban digitalization actually need?
To answer "what kind of network we need," it might be better to first figure out a more fundamental question: what exactly does city-wide intelligence include?
The essence of "city-wide" is to break down silos and treat the city as an organic "life form." Starting from this definition, the role of the network becomes clear. It must accomplish three things: enable city-wide perception through IoT devices, achieve high-speed circulation of data elements, and realize efficient scheduling of computing resources.
To achieve this goal, it is necessary to integrate scattered networks, plan and build them uniformly, and make them infrastructure just like water, electricity, gas, and roads.
In the past two years, with the maturation of technologies such as all-optical networks, intrinsic security, and AI operations and maintenance (AIOps), the industry has finally found a relatively systematic answer. The "One Intelligent Network" proposed by Huawei at the 9th Digital China Summit is exactly the representative solution in this round of exploration.
Its core can be summarized as "one network, dual intelligence": integrating scattered networks into "one network" and then endowing this network with the ability to think.
The so-called "one network" refers to the overall planning mentioned earlier, unifying the standards and planning of various urban networks to achieve city-wide coverage, while "dual intelligence" points to the business and technical levels respectively.
At the business level, the network needs to possess two core capabilities: intelligent connectivity and intelligent security.
The first intelligence in "dual intelligence" is to let the network "understand" the business. Traditional networks treat all business traffic equally without distinction. However, in the real scenarios of urban digitalization, the demands of different businesses vary drastically: video requires stable large bandwidth, industrial control requires ultra-low latency, and government data requires the highest level of security protection. If the network cannot identify these differences, it will either waste resources or cause critical business failures. What "intelligent connectivity" does is to enable the network to proactively identify each type of business and then allocate appropriate network resources for different businesses, just like an urban traffic system.
The second intelligence is to endow the network with an inherent security mindset. The traditional security approach is to build walls at the network boundary—firewalls, intrusion detection, layer by layer. But as the network expands and tens of thousands of terminals are connected, the boundary becomes increasingly blurred, and walls will eventually be breached. Attackers' targets are shifting from servers and applications to the network devices themselves. Once a router or switch is compromised, the entire network becomes useless.
Huawei's intrinsic APT (Advanced Persistent Threat) security solution embeds security capabilities directly into the network devices themselves, monitoring abnormal behaviors in memory, processes, and traffic in real time. It adds an intrinsic high-security board in the router, built with a billion-level threat signature sample library. Relying on the AI multi-modal protocol expert model, it can comprehensively identify abnormal behaviors in system files, processes, and memory.
When a suspicious process attempts to start, or abnormal traffic attempts to establish a covert channel, the system can complete identification and blocking within milliseconds through the "white-and-black" training model. More crucially, this system can link with the RDR² system to achieve 100% detection of unknown threats. For government networks carrying massive sensitive data, this intrinsic security capability is more reliable than any external firewall.
At the technical level, the network is also undergoing an upgrade in thinking capability—intelligent network elements and intelligent networks. Intelligent network elements refer to network devices possessing the capability to support native AI, enabling devices to identify businesses more accurately and provide more deterministic experiences for businesses. Through intelligent networks, network management and maintenance evolve from automation to intelligence, thereby consolidating the connectivity foundation in the intelligent era.
However, no matter how intelligent the connectivity is, it cannot be suspended in the air. What is the physical foundation that truly grounds it? The answer is the all-optical network.
The business characteristics of urban digitalization dictate that only optical networks can simultaneously meet the stringent requirements of large bandwidth, low latency, high security, and high reliability. A single optical fiber can carry multiple dedicated business networks simultaneously, with each physically isolated and free from mutual interference; quantum encryption technology provides theoretically "unbreakable" protection for highly sensitive data such as government affairs; the "1-millisecond latency circle" realized by the urban optical network allows computing power calls to be completed at the speed of light; and the intelligent perception of optical cables and the intelligent recovery capability of the optical network can achieve "broken fiber but unbroken network."
Taking the practice in Wuhan as an example, the local area has built "One Urban Network," creating a unified all-optical foundation with 100G ultra-broadband, achieving full coverage of the backbone network. Thousands of citizen requests, 220,000 video signals, and over 10 billion IoT data entries every day can flow into the urban operation management hub in real time. The network-wide bandwidth has increased by 10 times, the video service access speed has improved by 80%, and the unified optical foundation has achieved single-network carrying, reducing network usage costs by 25%. Currently, more than 30 cities nationwide have launched similar practices.
From the experience of these pioneers, a "thinking" network is moving from concept to reality. But a deeper question also emerges: why do we need to spend so much effort reconstructing the city's network foundation? What kind of changes will it ultimately bring?
When we zoom out, we will find that the proposal of "One Intelligent Network" comes at a critical turning point in urban digital transformation.
Looking back at history, urban digital and intelligent transformation has experienced a clear evolutionary trajectory.
In the "Digital City" stage from 2009 to 2014, the core was the construction of information infrastructure and e-government construction in single fields. In the "Smart City" and "New Smart City" stages from 2014 to 2023, emphasis began to be placed on systematic planning, data-driven approaches, and people-oriented development.
And the "city-wide digital transformation" since 2024 emphasizes treating the city as an organic life form. Through city-wide perception by IoT devices, high-speed circulation of data elements, and efficient scheduling of computing power, it realizes the full-scenario, all-around, and full-process intelligence of urban public services, urban governance, and digital economy businesses.
Under this definition, the role of the network has undergone a fundamental shift.
In the past, we understood the network as a "pipeline," a channel for data from point A to point B. But in the context of urban intelligent agents, the network has evolved from a pipeline to meridians. Although meridians do not bear weight directly, they run through the whole body and connect the organs, transmitting the information captured by perception to the computing power brain for decision-making, allowing data, acting as blood, to circulate and flow, and driving applications to execute tasks. Without meridians, organs act independently, and the nervous system has signals but cannot transmit them. The smoothness of the meridians determines the upper limit of the entire body's function.
In the AI era, businesses such as remote surgery and industrial control require deterministic experiences, which traditional networks are hard to support. "One Intelligent Network," through capabilities such as intelligent network elements, slicing technology, and deterministic networks, is upgrading best-effort networks into "deterministic" networks. This is a paradigm shift in network technology and a key prerequisite for the implementation of urban digitalization.
From the perspective of the industrial ecosystem, Huawei's "One Intelligent Network" has another strategic meaning: it proposes a new standard for urban network infrastructure. Currently, there is a widespread tendency in smart city construction to "emphasize applications while neglecting the foundation." Various regions are keen on launching various smart application platforms but neglect the overall planning and intensive construction of the underlying network.
The intelligent one network in the city is essentially calling for a mindset of infrastructure first: first connect, strengthen, and unify the city's network meridians, allowing data, acting as qi and blood, to reach the whole body without obstruction, and then grow various intelligent applications on the upper layer.
Returning to the question at the beginning: how can urban digitalization be both fast and stable?
"Fast" means making data run faster, computing power scheduling faster, and business response faster. The "1-millisecond latency circle" realized by the urban optical network makes millisecond-level computing possible and ensures the AI application experience. "Stable" means being stable in security, stable in reliability, and stable in sustainability. Intrinsic APT defense, quantum encryption, ASON (Automatically Switched Optical Network) protection... These capabilities build a network that can self-immunize and self-repair.
In the past, when we talked about smart cities, we talked more about large screens, algorithms, and platforms at the application layer. But more and more people realize that without a solid connectivity foundation, the upper-level intelligent applications are castles in the air.
The inspiration from Huawei's One Intelligent Network is: when AI begins to truly drive urban evolution, network infrastructure also needs to continuously evolve and upgrade synchronously on the basis of the past.
From acting independently to unified planning, from passive pipelines to proactive intelligence, from boundary protection to intrinsic security... This network is growing into a solid and smooth set of meridians, allowing the city to run both fast and steadily on the road of digitalization.