According to the latest data from the China Academy of Information and Communications Technology, China's AIoT market size exceeded 580 billion yuan in the first half of 2026 and is expected to cross the trillion-yuan threshold for the first time this year. In the public's intuitive impression, the Internet of Things has always been a hardware narrative of "sensors + communication modules + cloud platforms"—but an underappreciated fact is that revenue growth for pure software and services in this market has reached 47%, 2.6 times the hardware growth rate (18%). The Internet of Things is switching from a "connectivity economy" to an "intelligence economy," and the tech stack reconstruction at the software layer will determine the final distribution of this trillion-yuan pie.
Over the past five years, the industry's core narrative for the Internet of Things has been "number of connections": when the global number of IoT devices will surpass 50 billion, the growth curve of NB-IoT connections, and the coverage performance of WiFi6/Bluetooth Mesh. This logic barely held before 2023—hardware deployment was the prerequisite, and connectivity was value.
But entering 2026, the balance of drivers has clearly tilted. When the penetration rate of smart water meters exceeds 70% in some cities, and when annual shipments of smart door locks stabilize above 40 million units, the incremental space for hardware itself has already narrowed. What is truly raising the industry ceiling is the software layer running on every connected device: intelligent operations and maintenance on device management platforms, stream computing on edge gateways, and upgraded interaction experiences in user-facing mini programs or apps. In short:Hardware connectivity has completed "from 0 to 1," while software is doing "from 1 to N."
Take the smart community segment as an example: in a mature smart community solution, hardware costs (access control terminals, parking barrier controls, environmental sensors, video surveillance) as a share of total construction cost have fallen from 75% three years ago to 58% today. In their place, the value share of software modules such as property management SaaS, resident-facing mini programs, and facility operations and maintenance data analysis platforms continues to rise. This is highly similar to the development trajectory of the cloud computing industry—IaaS becomes thinner, while PaaS and SaaS become thicker.
If we set aside the grand narratives in industry reports and focus on the technical links that frontline developers actually encounter in real projects, the upgrades on the IoT software side are concentrated in three layers:
First, the maturation of edge computing middleware is rewriting the division of labor between device and cloud architectures.The standard paradigm for IoT development in the past was "device collection → MQTT reporting → cloud processing → command delivery," with long links and high latency, often falling short in video stream analysis or real-time alert scenarios. A change worth watching in 2026 is that edge containerization solutions based on lightweight Kubernetes distributions (K3s/MicroK8s) are beginning to enter production environments. A video analysis service deployed on an edge gateway can complete face recognition and abnormal behavior detection locally, syncing only structured events (rather than raw video streams) to the cloud. For software development teams, this meansthe need to simultaneously master embedded Linux trimming, edge inference framework deployment, and cloud data platform integrationthree skills, with development complexity rising rather than falling—but system response time compressed from seconds to milliseconds.
Second, the "modular assembly" trend in smart community management platforms is moving custom development from project-based work toward productization.Traditional smart community software development is typically project-based: each community conducts a requirements study, produces a customized plan, and delivers an independently deployed system. The problem is that a mid-sized property management company in Shenzhen manages 30 communities, and if each community requires customized development, operations and maintenance costs will devour all profits. The 2026 solution is to abstract subsystems such as security, access control, parking, energy consumption, and repair requests into standardized microservice modules, and use a configuration center to orchestrate differences by community. The front-end management backend combines pages in a low-code manner, while the resident side uses a unified mini program entry routed by community. This meansone core platform development can cover differentiated deployment across N communities—for the first time, the scale effect of software development has a practical path to implementation in the smart community field.
Third, lightweight interaction via device-side mini programs is replacing traditional apps as the preferred human-machine interface for the Internet of Things.An easily overlooked trend: for high-frequency IoT devices such as smart door locks, charging piles, and parcel lockers, more than 80% of users' first interaction is completed through WeChat mini programs rather than downloading standalone apps. Behind this is the result of user behavioral inertia—no one is willing to download a 200MB application just to unlock a door once. But for developers, mini program development means needing to complete integration of Bluetooth communication protocols, real-time synchronization of device status, and offline caching strategies in weak-network environments within a 2MB package size limit. Technically, this is harder than developing a native app becauseperformance optimization in constrained environments places higher demands on front-end engineering capabilities. The industry's current best practice is a combination of "mini program shell + Native plugin": core interaction and UI are completed at the mini program layer, while system-level operations such as Bluetooth/WiFi communication invoke native capabilities through plugins.
From a market rhythm perspective, the window period for the IoT software services layer is probably still 18 to 24 months. The judgment is based on three points:
First, leading cloud vendors (Alibaba Cloud IoT, Huawei Cloud IoT, Tencent Lianlian) are accelerating the construction of their respective device connectivity platforms and application development kits, but their focus is on the PaaS layer, and the SaaS application layer facing end customers still relies heavily on secondary development by ISVs and SIs. This division of labor will not change in the short term—cloud vendors do infrastructure, ISVs do industry applications, and a huge space for software development exists in between.
Second, industry standards have not yet been unified. Although the Matter protocol has made progress in the smart home field, in B-end scenarios such as smart communities and industrial IoT, proprietary protocols from various vendors still dominate. Standard fragmentation means demand for customized software development will persist for a long time; a single large platform cannot "win everything with one move."
Third, the application of large models on the multimodal perception side has only just begun. The current "intelligence" of AIoT is more reflected in single-point analysis—face recognition, license plate recognition, smoke detection—rather than joint reasoning across sensor types. Once multimodal large models can directly consume heterogeneous data streams from IoT devices, upgrading from "independent alerts from each sensor" to "comprehensive scenario understanding," the entire IoT software development tech stack will undergo another round of upgrades.
For practitioners in the software development field,the Internet of Things is no longer a "domain" that needs to be learned separately, but a "dimension" that will permeate all software systems. Whether doing management backends, mini program development, or data platforms, in the next two years there is a high probability of encountering IoT-related technical links such as device access, protocol adaptation, and edge deployment in some project.
In Shenzhen, more and more enterprises are beginning to integrate IoT capabilities into their own digital solutions—not separately doing an "intelligent hardware project," but adding a device management layer to existing business systems. This "+IoT" rather than "IoT-centric" approach may be the more actionable implementation path at present. After all, technology itself is not the goal,using technology to create value—making device data truly serve business decisions and user experience—is the ultimate proposition of the IoT software layer.
Room 1409, Building C, Shenzhen Century Industrial Center, No. 743 Zhoushi Road, Hezhou Community, Hangcheng Sub-district, Bao'an District, Shenzhen
