Author: Peng Zhao (Founder of Zhicifang, Co-Founding Partner of Yunhe Capital) Original
By the end of 2025, a figure quietly crossed a watershed: the proportion of "things surpassing humans" (where IoT connections exceed mobile phone connections) in mobile network terminal connections reached 61.3% (according to the "2025 Statistical Communiqué on the Communications Industry" released by the Ministry of Industry and Information Technology (MIIT)). The number of machine connections has not only completely surpassed that of humans but also accounts for over 60% of the total.
Most people treat it merely as a statistical figure, while I read it as a "curtain divider": the second act of the IoT (Internet of Things) is drawing to a close, and the third act is gradually unfolding.
This article does only two things:
First, conduct a comprehensive review of the second act using the four-layer framework of "Connectivity, Sensing, Intelligence, and Value." In late March this year, the MIIT and eight other departments jointly issued the "Action Plan for Promoting the Innovative Development of the IoT Industry (2026-2028)," explicitly setting the goal of breaking through the 3.5 trillion RMB mark for the core IoT industry scale by 2028. We will focus on dissecting how this 3.5 trillion is distributed and where the incremental growth lies in which layer.
Second, value the third act and explain why I believe the true incremental growth of the IoT may not lie within this 3.5 trillion at all.
In the current policy context, there are three easily confused trillion-level figures with completely different statistical calibers:
3.5 Trillion RMB: This is the 2028 target for the core IoT industry scale just mentioned, specifically referring to the scale of the IoT industry entity itself, including chips, modules, terminals, networks, platforms, and application services.
2.5 Trillion RMB Level: This is the 2030 policy target for the added value of the core industrial Internet industry (from the "Implementation Opinions on Promoting the High-Quality Development of the Industrial Internet" issued by the MIIT and seven other departments in June 2026; according to calculations by the China Academy of Information and Communications Technology (CAICT), the present value in 2025 is about 1.67 trillion RMB). It overlaps with the former but does not coincide.
7 Trillion RMB Level: This is the annual investment for the "Six Networks" (water conservancy, new power grid, computing power, new-generation communication network, urban underground pipe network, and logistics network, deployed at the Politburo meeting of the Central Committee in April 2026) in 2026. It is not the IoT industry scale, but the upstream funding pool on the demand side of the IoT.
According to calculations by the IoT Think Tank, the base period scale of China's core IoT industry in 2025 is about 2.6 trillion RMB. To achieve the 3.5 trillion target, the implied compound annual growth rate (CAGR) is about 10%. For an industry with such a massive base, this is not low. However, we will argue later that this is highly likely just a conservative starting point, and the ruler itself only measures the tip of the iceberg, because the other half of the story has just begun.
Perspective on the 3.5 Trillion Landscape Using the "Connectivity, Sensing, Intelligence, and Value" Framework
Traditional divisions such as "perception layer, network layer, platform layer, and application layer" are now inadequate in today's AI explosion.
As large models move to the edge, the boundary between platforms and intelligence is constantly being reconstructed (platforms are evolving from simple connection management to AIoT agent platforms), leaving operating systems and edge intelligence with no place in the old framework; more critically, the true commercial closed-loop of the IoT is often glossed over by the broad term "application," failing to reflect the real premium of data assets and scenario operations.
Therefore, to more accurately analyze the core value flow of the industry, we adopt a new four-layer framework of "Connectivity, Sensing, Intelligence, and Value" to re-measure the industry:
Connectivity (Connection Services): Cellular connections, 5G private networks, RedCap/satellite/passive and other new types of connections, counting only connection service fees, excluding equipment;
Sensing (Perception and Hardware): Chips, modules, SIM/eSIM cards and components, video hardware, industrial networking hardware, pan-AI hardware terminals (robots, AI toys, AI wearables), sensing components... full hardware caliber;
Intelligence (Platform and Intelligence): IoT platforms, operating systems, edge intelligence and AI enablement, IoT security;
Value (Value Realization): Industry solutions, scenario operations, and data value-added services, aggregated into four industry groups: digital and intelligent production, governance, life, and long-tail.
The first structural conclusion measured by this new ruler is that hardware (the "Sensing" layer) accounts for about 46%, making it the largest single layer, but the "Value" layer for value realization constitutes the largest single incremental direction at about 42%. That is: hardware is the entry point, and value lies in applications. This structure is more robust than any total volume forecast and is also the key to understanding the trajectory of the IoT during the 15th Five-Year Plan period.
To panoramically dissect the specific scale of these four layers, the IoT Think Tank provides two forecast scenarios: baseline and challenge.
The baseline scenario aligns with the pace of official targets (CAGR of about 10%~11%); the challenge scenario (CAGR of about 14%) is the landing point we believe is more likely to be achieved. Under both scenarios, the proportion structure of the four layers remains stable.
Estimation Table of the Scale of China's Core IoT Industry (Connectivity, Sensing, Intelligence, and Value Framework)
Key points and public anchors for each layer:
Connectivity (about 4%~5%): Small pie, big leverage. By the end of 2025, "things surpassing humans" accounted for 61.3% of mobile network terminal connections; extrapolating from the target of 3.6 billion connections for "Intelligent Connection of Everything" by 2027, cellular connections will reach about 4 billion in 2028. Although the proportion of connection services is small, it determines data ownership and platform entry points; its strategic weight is far greater than its revenue weight. Future incremental growth will focus on three areas: the opening of 5G standalone private network pilots, the scaled deployment of RedCap, and satellite direct connection and passive IoT.
Sensing (about 46%): Largest single layer, with the greatest elasticity. Chinese manufacturers hold an absolute advantage in global shipments in the cellular module market; video hardware is anchored by the financial reports of leading listed security companies. But the real variable is pan-AI hardware terminals (AI glasses, AI toys, AI companions, robots). According to public reports, overseas giants have given aggressive supply-side signals, with the first companion hardware from leading AI companies targeting shipments of hundreds of millions of units, and the capacity planning for AI glasses from social media giants reaching the tens of millions level. Under the challenge scenario, the "Sensing" layer will surge to the 2.3 trillion level by 2030, with the main elasticity coming from this.
Intelligence (about 8%): Small proportion, but highest growth rate. Policies explicitly state that by 2028, there will be over 450 industrial Internet platforms and more than 120 million connected industrial devices, and propose to build 5 comprehensive platforms with international influence by 2030. Platforms are evolving from "connection management" to AIoT agent platforms of "platform + large models." Note that we have mostly included the scenario-side monetization of platforms in the "Value" layer; this layer only counts the platform entity, OS, edge intelligence, and security. Do not underestimate the strategic position of platforms due to their low proportion.
Value (about 42%): Largest increment, funds are already on the way. The certainty on the demand side comes from the funding end. The annual investment in the "Six Networks" exceeds 7 trillion RMB. Estimating based on the proportion of sensing and networking equipment and systems in new infrastructure investment being 1% to 3%, it can be transformed into 70 billion to 200 billion RMB of direct IoT demand annually. This does not even include the strong pull from equipment updates and consumption upgrading under the "Two Major, Two New" policy.
Three Accelerators Drive "Better-than-Expected" Growth
Accelerator One: The supply-side super cycle of edge AI. 2026-2027 is a window for intensive launches of pan-AI hardware. Every pan-AI terminal drives the entire chain of "connection + chip module + platform invocation + content services," and a significant proportion will adopt cellular or eSIM connections, which is a structural dividend not available in the previous wave of smart speakers.
Accelerator Two: The concentration of funds on the demand side. The 7 trillion RMB for the "Six Networks," the "Two Major, Two New," and "consumption upgrading" are three streams of funds superimposed in the same direction, and the IoT is their common "sensing foundation" procurement item. Historically, the IoT has never received such a high-density demand-side funding arrangement within a single five-year period.
Accelerator Three: Resonance of policy and technology readiness. The "AI+" action explicitly sets a 70% penetration rate by 2027, the cost of RedCap chip modules drops to the inflection point for scaling, eSIM resumes commercial use on the terminal side, and satellite direct connection moves from technical verification to commercial operation. The readiness windows for these four technologies rarely overlap in the same five-year period.
Therefore, based on comprehensive judgment, the baseline scenario is the prudent lower bound, and the challenge scenario is closer to our expected landing point. The 3.5 trillion target is highly likely to be achieved ahead of schedule or exceeded between 2027 and 2028. Mainstream market forecasts underestimate the category creation effect of edge AI on the "Sensing" layer and the transmission speed of funds from the "Six Networks" to the "Value" layer.
This structural change also points out the subsequent strategic focus for various participants in the industrial ecosystem.
For hardware manufacturers, the 46% pie is the foothold, but homogenization will be faster. The real value of hardware lies in the "entry point"; a single module determines connection and platform affiliation. According to industry experience, every 1 RMB of hardware revenue can drive 3 to5 RMB of connection, platform, and service revenue. The mindset of selling hardware cannot scale up; only the mindset of building entry points can capture the 54% outside the 46%.
For platform-oriented enterprises, they need to accept the destiny of becoming "foundational," with the main battlefield for commercialization on the application side. Fighting for a national-level sequence seat among the "5 comprehensive platforms" is much more important than simply fighting for platform subscription fees.
For solution providers and operators, the 42% "Value" layer is the largest increment, and the funding context is clear. Mapping product lines one by one to the funding accounts of the "Six Networks" and the "Two Major, Two New" is much more effective than broadly talking about digital transformation.
Finally, a universal estimation tool for all enterprises to evaluate their opportunities in any segmented market: use the three-factor method of "Addressable Space = National Scale of the Segment × Product Relevance × Maximum Share," calculating bottom-up for each product line and then summing them up.
Looking one step further ahead, the business model of the IoT is undergoing its third transformation: selling connections (SIM and traffic) → selling capabilities (PaaS/SaaS) → selling results (AIoT agent as a service, A-AaaS). When intelligence is measured uniformly in Tokens and billed by business results, the billing logic of the entire industry may need to be rewritten.
Arrow Reversal: When Intelligence Gets a Body
Now, let us return to the "curtain divider" at the beginning and put the IoT into a larger narrative.
Act One, Connection: Connecting things to the network, giving the atomic world a household registration in the bit world for the first time. The protagonists are operators, and the unit of measurement is the number of connections.
Act Two, Perception and Data: Sensors, cameras, and modules move the physical world into the bit world in an organized manner; the direction of the arrow is "atoms → bits." The protagonists are platforms and data. "Things surpassing humans" is the closing bell for this act. When machine connections account for over 60%, the "digital relocation" of the physical world is mostly complete, and the dividend of simple connections has ended.
The hallmark of Act Three is the arrow reversal: bits → atoms. Intelligence is no longer satisfied with observing and analyzing the physical world in the cloud; instead, it gets a body and reverses into the physical world to execute and transform. Robots, AI glasses, AI toys, and autonomous driving carriers are all the vanguard of this reverse migration.
Seeing this clearly, one will understand why pan-AI hardware terminals are not just "smarter things." The subject of the first two acts is "things," the things of the IoT; in the third act, the subject changes to "intelligence." Devices are no longer objects to be connected, but physical carriers of intelligence. With the subject changed, the "grammar" of the industry changes entirely.
We call this approaching stage AIoT 3.0.
The native species of the third act also has its own scientific name: AIoT Agents. It is a software-hardware integrated intelligent form that implants perception, computing, and execution capabilities into physical devices, simultaneously being a "citizen" of the bit world and a "resident" of the physical world. Any device that can perceive, make decisions, and exert actions on the physical world is completing "AIoT agentification."
Translated into industrial language, the development of the IoT is a relay of verbs. In the past, we experienced the upgrade from "Internet of Everything" to "Intelligent Internet of Everything." At last year's AIoT Industry Annual Conference, we proposed "Actionable Internet of Everything" (万物智行), declaring that AIoT agents will act as an independent "species," autonomously perceiving, deciding, acting, and even trading in both the physical and digital worlds.
A year has passed, the arrow has reversed, and agents are beginning to comprehensively take over the execution and transformation of the physical world. The direction declared last year is growing into a complete era.
From this comes the most important judgment of this article: the 3.5 trillion may only measure the scale of the second act. The official statistical boundaries (chips, modules, terminals, networks, platforms, application services) are products of the "atoms → bits" era. After the arrow reverses, the output value created by intelligence executing and transforming the physical world—the work done by robots, the production lines reconstructed by embodied intelligence, the assets operated by agents—will mostly fall outside this statistical caliber.
Therefore, the correct attitude towards the 3.5 trillion is to treat it as a phased answer sheet for the second act, and then realize that the true increment may not even be in this answer sheet. An easily overlooked point is that the ruler for measuring industrial value itself is probably about to be reconstructed, which is the largest expectation gap facing the industry during the 15th Five-Year Plan period.
Connection, Intelligence, Action—the relay of verbs has run to the third leg. But a complete era, besides verbs, also needs a name of its own. What is it? At the AIoT Industry Annual Conference on November 27 this year, we will officially unveil it.
Final Thoughts
From the Internet of Everything to the Intelligent Internet of Everything, and then to the Actionable Internet of Everything, the past few decades of the IoT have been a continuous relay of verbs.
"Things surpassing humans" declares that the climax of the second act has passed; the four-layer framework of "Connectivity, Sensing, Intelligence, and Value" for the 15th Five-Year Plan and the 3.5 trillion industry target are a serious review of the second act. While reviewing, let us also look up and see the road ahead together: intelligence is getting a body, stepping into the physical world, and the curtain of the AIoT 3.0 era is opening. As for the exclusive name of this era, we will see it clearly on November 27 in Shenzhen.
The arrow has reversed, and the third act is beginning.
Note:
The cited data is classified as follows:
The first tier consists of public data from official and authoritative institutions such as the MIIT, the National Bureau of Statistics, and the CAICT;
The second tier consists of public calibers from international research institutions such as Counterpoint and IDC;
The third tier of market institution data is only used for magnitude reference.
The enterprise information involved in the text is all from public channels (listed company annual reports, public reports).