The 9th China International Import Expo will be held from November 5 to 10 at the National Exhibition and Convention Center in Shanghai. As the world's first national-level expo with imports as its theme, this year's Technical Equipment exhibition area has set up a dedicated zone for "digital industrial automation," and the Lingang exhibition area has also expanded to 2,700 square meters, bringing together more than 100 enterprises and over 150 debuts and first-time showcases. One clear signal is this: the narrative of industrial AI is shifting from "being able to chat and answer questions in the showroom" to "being able to get work done on the production line."
For Chinese manufacturing enterprises and CIOs responsible for digitalization and intelligent manufacturing, this is not just hype—it is a question that must be answered: when industrial large models truly land at core workstations such as welding, quality inspection, and process optimization that were previously difficult to standardize, are your ERP, MES, and PLM ready to catch it?
As of June 2026, mainland China has 103 lighthouse factories, accounting for 43.3% of the global total (Xinhua News, 2026-09)
I. CIIE Exhibition Signals: From 'Single-Point Device Intelligence' to 'Full-Chain Digital Native'
The key exhibits pre-revealed for the "Digital Industrial Automation Zone" in the CIIE's Technical Equipment Exhibition Area almost all point in the same direction—deeply integrating digital technology with industrial equipment and production processes to achieve automatic sensing, decision-making, execution, and optimization. A few representative examples:
| Exhibitors / Exhibits | Core Competencies | Implications for Manufacturing Digitalization |
|---|---|---|
| Rockwell iTRAK Flexible Manufacturing Line + Emulate3D | Flexible production line + dynamic digital twin simulation | Production line design and operation close the loop within the same digital mirror |
| Jungheinrich SMART Plus 3.0 | Digital Twin + AI Intelligent Decision-Making | Equipment utilization 92.1%, energy consumption -18.6% |
| Johnson Controls YORK® YMSC | Air-suspension variable-frequency centrifugal chiller | Save up to 50% on annual operating costs |
| Tesla (Debut at Lingang Exhibition Area) | Cybercab / humanoid robot / energy storage | Concentrated showcase of intelligent manufacturing and energy AI layouts |
The common feature of these exhibits is that they no longer showcase a single "smarter machine tool" on its own, but rather a digital chain that runs through "design—simulation—production—operation and maintenance." This aligns with the judgment at the 2026 Shanghai Industrial Fair—manufacturing digitalization is moving from "single-point equipment intelligence" to "full-chain digital native": one full-chain digital solution for automotive parts compressed the development cycle for new parts from 18 months to 6 months.
The turning point of industrial AI is not that "the models have become stronger," but that "the main data line has been connected." The prerequisite for models to be deployed at workstations is that data from design, process, manufacturing, and quality flow along the same line.
II. Digital Thread Connecting PLM→ERP→MES: The Schneider-PTC Logic Is Becoming Consensus
Why has the Digital Thread suddenly become a keyword? Because it solves manufacturing's most painful long-standing problem: design in PLM, costs and planning in ERP, and processes and actual performance in MES have long been three mutually disconnected networks. Schneider's all-cash acquisition of PTC for $22.6 billion (disclosed in 2026-10, at $205 per share, a 42.3% premium, expected to close in Q3 2027) is precisely aiming at this full-lifecycle data thread from design→build→operate→maintain. PTC has 30,000+ customers; after the transaction is completed, software and services will account for about 24% of Schneider's revenue.
This is not an isolated case. Sailun's “Rubber Chain Cloud” has connected 150 system interaction interfaces, linking more than 3,200 suppliers, more than 2,000 distributors, and nearly 60,000 stores; the China Resources Sanjiu Guanlan Base in Shenzhen relies on digital twins and WMS automated warehouses to achieve full-process traceability, while Sharetronic Data's Yantian factory deeply integrates MES, WMS, and AI large models to automatically generate optimal production plans based on historical sales and market trends. The logic is highly consistent: whoever weaves the digital thread first gets the ticket to industrial AI implementation.
Isolated island
Traditional MES/PLM/ERP systems each operate in their own silo: change one thing in design and the process must be re-sequenced, cost figures don't line up, and AI models can't get consistent context — so they never get past the demo stage.
Main line
Digital thread continuity: the same product definition runs through R&D—planning—manufacturing—service; only when AI obtains trusted data at the workstation level can autonomous optimization and prediction be discussed.
III. Industrial Large Models Deployed at Production Workstations: The Break in the Data Mainline Is the Real Bottleneck
Another signal from the 2026 Shanghai Industrial Fair is critical: industrial large models are no longer just Q&A demos on exhibition booths, but are directly embedded into core production workstations such as welding, quality inspection, and process optimization. But implementation easily runs into pitfalls—many enterprises have bought industrial large models only to find that "the model is smart, but the production line doesn't buy it." The root cause is often not in the model, but in a broken data mainline.
This is isomorphic to the open-source data foundation direction we have previously tracked: the OPC Foundation's open-source Cloud Initiative reference solution uses MIT-licensed open components to list MES/DCS/ERP as integration targets, enabling enterprise data flows without changing existing OT systems, emphasizing "Zero lock-in" and multi-cloud compatibility, and responding to the EU Data Act and the Digital Product Passport (DPP). Only when the data mainline is connected through open standards rather than proprietary interfaces can industrial AI have stable, auditable "ingredients."
Three typical manifestations of data mainline disruption:
① Semantic inconsistency: The same material has different codes in PLM, ERP, and MES, so AI cannot confirm "it refers to the same thing".
② Lack of real-time capability: Process parameters are synchronized through end-of-day batch processing; the model sees "yesterday's production line," so decisions are inherently delayed.
③ Unclear responsibilities:AI-generated process recommendations are not reviewed by anyone and leave no records, so when quality issues occur, they cannot be traced to the specific point.
IV. Practical Recommendations for Manufacturing Digitalization Leaders
1. First establish "data interoperability standards," then deploy industrial large models. Prioritize using standards such as OPC UA and open APIs to connect MES/ERP/PLC data, avoiding lock-in by a single vendor's proprietary interfaces. Only when the standards are connected can models have trustworthy context.
2. Choose the PLM–ERP–MES integrated path. It is not necessary to build everything in one step; you can start from the data mainline across the three stages of "design—planning—manufacturing," verify the value of end-to-end connectivity, and then expand to service and operations.
3. Industrial large models start from small scenarios such as "quality inspection/process". Data at these workstations is relatively clean and its value is quantifiable, making it easy to achieve ROI; don't start by challenging strongly constrained, high-risk areas such as production scheduling.
4. Governance and security are on the table at the same time. AI process recommendations must be traceable, reviewable, and auditable, meeting the compliance requirements for private deployment, Classified Protection, and Xinchuang — innovation cannot bypass risk control.
An often overlooked reminder: the digital thread is "slow work," with no explosive KPIs in the short term. But it determines the ceiling of your industrial AI over the next three years. Weave the thread densely, and models are merely the icing on the cake; if the thread is broken, even the strongest model is just a performance in a showroom.
The CIIE opens tomorrow, and the spotlight of industrial AI will shift from "concepts" to "workstations." For manufacturing enterprises, the real race is not about who is the first to shout the slogan of "industrial large models," but about who is the first to weave the data thread among PLM, ERP, and MES into something dense, connected, and trustworthy. This thread is the moat of the industrial AI era.
Data sources for this article: the official website of the China International Import Expo (the 9th CIIE, 2026/11/5–10; "Digital Industrial Automation Zone" in the Technical Equipment Exhibition Area, 2026-10-09; Lingang Exhibition Area, 2026-10-19, ciie.org); Xinhua Net's "Super Factory: Transformation with Method" (2026-09-15); interpretation of the 2026 Shanghai Industrial Fair (ciif-sh.org.cn); this series' reports on Schneider's acquisition of PTC (2026-10-22) and the OPC open-source data layer (2026-10-20); Shenzhen Municipal Government policies related to smart factories (sz.gov.cn, 2026). Data as of 2026-11-04.
Zhongdi Information · Dedicated to Odoo and digital solutions for enterprise management
