The W2 Hall Metaphor: Why Video Surveillance Is No Longer the Star of Security China

A floor plan is an industry statement that never speaks. AI+ took the E1 core hall while intelligent video surveillance was moved to W2 — not because it stopped mattering, but because it no longer carries the story on its own.

Security China 2026 — the 18th China International Exhibition on Public Safety and Security — closed in Beijing on 6 November. Most post-show coverage understandably focused on spherical robots, counter-drone systems and large models. This six-part series reads the same event from a different vantage point: what it changes for the people who specify, build and buy in-vehicle video systems.

The Floor Plan Is the Argument

The show was laid out as one core hall plus five specialist halls: E1 for international exhibitors and the AI+ pavilion, E2 for smart policing and digital security, E3 for low-altitude security and intelligent robots, W1 for digital cities and smart living, W2 for intelligent video surveillance, and W3 for smart access and alarm systems. AI and embodied intelligence occupied the core; video surveillance was assigned a specialist hall.

The organiser listed five headline themes in order: AI large models and security agents, a first-ever embodied intelligence core zone, low-altitude security, smart policing, and cross-domain integration. Intelligent video surveillance was not among them. The official innovation and solution shortlist gives harder numbers: 158 companies filed entries, with 103 innovative products and 87 solutions selected.

Three Moves Hidden in a Map

For two decades the camera was the undisputed protagonist of this show. The industry argued about resolution, wide dynamic range, low-light performance and codec efficiency; the value of a device was almost equal to its imaging and recording capability. This edition inverted that logic.

The first move is from device to compute and model. E1 was named the AI+ pavilion and its exhibit list covered large models, agents, intelligent terminals, compute algorithms and key chips. Huawei, Huawei Terminal, China Telecom, Tiandy, Vimicro, Fullhan, VSigma and Goke Microelectronics presented security-industry large models, public-safety agents and multimodal perception algorithms there.

The second is from seeing to acting. Wang Nan, secretary-general of the China Security and Protection Industry Association, put it plainly: security used to solve the problems of eyes and ears; now, by enriching the brain and adding the ability to act through robots, embodied intelligence is reshaping the entire industry. In a pre-show interview he described the sector as entering a phase defined by digital intelligence, convergence, unmanned systems, compliance and internationalisation, and specifically named on-device and edge deployment as new growth space.

The third is the least noticed: from whole units to components. Read the video surveillance category of the official shortlist and complete cameras are scarce. Most seats went to silicon and parts — Hisilicon VLM multimodal camera chips and a domestic codec SoC, Goke black-light AOV vision chips, VSigma on-device AI chips, a reconfigurable TPU, Biwin video-surveillance eMMC, WD Purple Pro 26TB drives and a fully domestic embedded motherboard. Imaging, storage, compute and encoding have become separately purchasable capability units.

Our Take

The camera was not eliminated; it was demoted to substrate. Once models, silicon and command platforms carry the narrative, a camera stops being a solution and becomes a sensor inside someone else's solution. For any vendor whose core asset is a hardware box, that is the same warning: if differentiation is left to channel count, resolution and price, you will eventually be compressed into a single line on a bill of materials.

In-vehicle surveillance is especially exposed. An MDVR arrives with a black-box label — customers buy it to satisfy an archival obligation, not because it creates decision value. Once vehicle makers specify intelligent video monitoring as factory-fit standard equipment, even that baseline demand can be taken over by the OEM layer.

Who Took the Vacated Space

Three of Wang Nan's six opportunities for the next five-year period are worth copying into a product roadmap: on-device and edge deployment, embodied intelligence and low-altitude unmanned systems growing faster than traditional security business, and the upgrade of the installed base plus data-value release and domestic substitution. He did not say more cameras; he said installed base and data as a factor of production.

The policy framing agrees. The State Council AI+ action opinion (Guofa 2025 No. 11) targets over 70 percent penetration of new intelligent terminals and agents by 2027, explicitly strengthens the role of AI in public governance, and calls for a model capability evaluation system plus Model-as-a-Service and Agent-as-a-Service. The show organiser wrote that same 2027 goal into its mission statement. The floor plan is less an aesthetic choice than industrial policy projected onto an exhibition map.

Our Judgment

Over the next three years, in-vehicle surveillance vendors can defend three positions; a fourth will be squeezed out. One: on-device intelligent hardware, turning compute, imaging and power into automotive-grade barriers. Two: a scenario data closed loop, turning alerts into evidence that is attributable, actionable and reviewable. Three: a trusted and compliant foundation, making video hold up legally and across borders. Four — pure recording hardware integration — faces commoditisation, OEM squeeze and a price war.

What to Take Away

• Redraw your product line around which layer of value the customer pays for, not around channel counts or TOPS.

• Read the official shortlist as a supply-chain map: chips, storage, encoding and security components are now domestically available at every layer.

• Watch 2027. The 70 percent penetration target for intelligent terminals and agents is the top-level budget driver behind procurement.

Sources: organiser theme interview and innovation shortlist, China Daily, CNR, China Industry News, the State Council AI+ action opinion, and the published hall plan. Exhibition scale figures conflict across official pages, a government calendar, a tender notice and closing reports; this article deliberately makes no precise claim about square metres or visitor counts. All vendor parameters are manufacturer claims, not independent test results.