AIoT Enters a New "Intelligent Connection" Phase: From Connecting Everything to Intelligently Acting

At the 2021 Shenzhen AIoT Summit, topics such as in-memory computing chips, AIoT platforms, and energy IoT were still frontier explorations. Four years later, the industry has undergone a fundamental paradigm shift. China's AIoT solutions market grew from approximately 54.1 billion yuan in 2020 to 111.9 billion yuan in 2024, with a compound annual growth rate of nearly 20%, and is projected to reach 128.7 billion yuan by 2025. More notably, the nature of the technological path has changed—AIoT is moving from "connecting everything" to "intelligent connection" and even "intelligent action."

Substantive Breakthroughs in Edge Intelligence

If the AIoT discussion in 2021 was still at the cloud-edge collaboration stage of "AI as the brain of IoT," the core trend today is the comprehensive下沉 of intelligence to the edge and terminals. At the 2025 World IoT Expo, Tang Hao, deputy director of the Wuxi IoT Innovation Promotion Center, used a precise metaphor: "In the past, IoT collected all data, like putting the entire ocean into storage, while intelligent sensing only needs to extract the key 'drop of water.'" A machine tool sensor now only initiates data transmission when an anomaly is detected, remaining "silent" during normal operation—behind this is a substantive improvement in edge computing capabilities.

This change is more intuitive at the hardware level. Six-axis force sensors now integrate computing chips and dedicated algorithms, giving industrial robotic arms and humanoid robots intelligent joints capable of "thinking independently." Li Dalong, product marketing director at Qualcomm Technologies, pointed out at the 2025 Shenzhen International Electronics Show that "always-on connectivity at the edge, AI-empowered edge computing, and AI-aware workflows" are the three major catalysts for this round of change. Edge AI is no longer a concept but is redefining the capability boundaries of terminal devices.

City-Level "Thinking" Capability

The second significant change in AIoT is the shift from device intelligence to system intelligence. More than 4,000 cameras and traffic lights in Wuxi are all connected to the same network, achieving "integrated global signal control," and the city's traffic system has begun to "learn" how to dispatch traffic flows. Some connected vehicle lanes can "deform"—borrowing the opposite lane for left turns during morning rush hour, and tidal robots moving isolation barriers during evening rush hour, allowing lanes to switch freely between tidal flows. Buses enjoy green-light priority, and autonomous taxis travel through streets and alleys.

This city-level intelligent collaboration was unimaginable at the 2021 forum. At that time, discussions focused on intelligent transformation of single scenarios, while today's AIoT systems can span multiple subsystems and achieve dynamic optimization at a larger scale. Shenzhen's Longgang District, guided by the "AI in AI" strategy, is also exploring a dual-engine driven model of governance and industry.

Security Paradigm Faces Reconstruction

At the 2021 forum, security topics mainly revolved around basic cryptographic algorithms of security chips. Today, the security challenges of AIoT have undergone structural changes. With the普及 of edge intelligence, the attack surface has expanded from the cloud to every terminal device with inference capabilities. AI itself is being used for real-time threat detection—CNN models achieve 90.7% accuracy in IoT malicious traffic detection, a 42% improvement over traditional Snort rule systems.

More transformatively, zero-trust architecture is merging with AI, blockchain, and edge computing. Traditional centralized security models are no longer sufficient to cope with dynamically changing attack vectors, and the principle of "never trust, always verify" combined with real-time computing capabilities at the edge is building a more resilient security framework. Security is no longer an "add-on" but a foundational capability embedded in every layer of the architecture.

Shenzhen's Role Transformation

At the 2021 forum, Shenzhen appeared more as a forum venue and industrial cluster. Now, Shenzhen is becoming one of the definers of AIoT industry rules. In 2025, Shenzhen's core AI industry revenue reached approximately 220 billion yuan, with more than 2,600 typical enterprises gathered. IOTE 2025 Shenzhen IoT Expo brought together 1,001 companies, with 118,000 visitors over three days, including overseas visitors from more than 30 countries and regions.

More strategically, Shenzhen has begun systematic布局 in segmented tracks such as humanoid robot sensors and edge computing. The "Humanoid Robot Sensor Industry Patent Navigation Report" initiated by the Shenzhen IoT Industry Association attempts to provide decision-making support for future industries from the dimensions of technology route evolution and industrial competition landscape. This role transformation from market follower to rule participant is one of the most fundamental changes in the past four years.

From "Connection" to "Action"

Looking back at that 2021 forum, the core proposition was "the integration of AI and IoT." Four years later, integration has been completed, and the new proposition has become "how intelligent agents can act effectively." When IoT terminals become the "senses" through which AI perceives the world, and AI algorithms become the "brain" of IoT, the competitive focus of the industry has shifted from "can it connect" to "can it do the right thing."

This shift means that the value assessment criteria for AIoT are moving from connection numbers and data volume to decision quality and action effectiveness in actual scenarios. Whether energy-saving algorithms truly reduce energy consumption, whether predictive maintenance reduces downtime, whether intelligent scheduling alleviates congestion—these "last mile" effects will determine the quality of the next growth cycle of the AIoT industry.