2025 China Visual Inspection Industry: From "Encouraging Machine Vision" to "Mandatory Inspection Capabilities"

I. Why Has "Visual Inspection" Suddenly Become a Focus of Capital, Industry, and Policy?

1. Capital Side: Second Only to Humanoid Robots in Popularity

In 2024, the number of publicly disclosed financing events in the industrial vision field ranked second among the advanced manufacturing sub-sectors, second only to humanoid robots; however, the average financing amount per deal was higher. Zhongyan Puhua Capital Group's tracking revealed that funds are concentrating at both ends:

- Upstream core components (high-end CMOS, 3D sensors, AI accelerator cards);

- Downstream vertical integration solutions (power batteries, photovoltaics, semiconductors).

"Middle-layer" pure system integrators, without algorithm platforms or industry data accumulation, generally face valuation pressure.

2. Industry Sector: New Energy + Semiconductor "Dual-Wheel" Growth The photovoltaic TOPCon and HJT new routes require increasingly higher precision in detecting microcracks and coating defects, moving from micrometer-level to sub-micrometer-level. In the semiconductor front-end and advanced packaging stages, domestic equipment manufacturers must develop their own vision inspection modules to secure entry into the fab manufacturing sector. According to statistics from the China Research & Consulting Group, the vision inspection orders released by the new energy and semiconductor industries alone in 2025 will be equivalent to the entire market size in 2020.

3. Policy Sector: From "Encouraging Machine Vision" to "Mandatory Inspection Capabilities" The Ministry of Industry and Information Technology's "Implementation Opinions on Improving the Reliability of Manufacturing" for the first time proposed the hard target of "100% online inspection of key processes." The State Administration for Market Regulation implemented a pilot program of "mandatory recall of defective products" for lithium batteries and photovoltaic modules, forcing companies to deploy vision inspection at the factory stage. More importantly, the second phase of the National Integrated Circuit Industry Investment Fund included "core components of industrial vision" in its key support catalog, placing them on par with semiconductor equipment and components. This signifies that vision inspection has been upgraded from a "tool" to a "strategic security link." II. Competitive Landscape: International Giants "Defend High-End," Domestic Brands "Seize Scenarios"

1. Upstream: High-end Components Still Face Bottlenecks, But Domestic Substitution Reaches a "Critical Point"

- High-end CMOS: Sony and ON Semiconductor still hold 70% of the global market share. Domestic brands like Will Semiconductor and Galaxycore have approached their counterparts in key indicators such as dynamic range and readout noise, and are expected to see significant volume growth in "non-fab" scenarios such as power batteries and photovoltaics by 2025.

- 3D Sensors: Keyence and LMI lead the way in structured light and line laser technologies. Domestic manufacturers are capturing emerging scenarios like lithium batteries and heavy-duty truck battery swapping through "self-developed core algorithms + localized services," nearly doubling their market share in three years.

- AI Accelerator Cards: NVIDIA's industrial-grade GPUs are in short supply and expensive. Huawei Ascend, Cambricon, and Horizon Robotics are entering the market with integrated solutions of "chip + framework + visual algorithm library," rapidly increasing customer loyalty.

2. Midstream: The "Smile Curve" of System Integration Intensifies Differentiation

Zhongyan Puhua Industry Research Institute categorizes system integrators into three types:

- Hardware OEM: Purchases foreign cameras and lenses, performs simple assembly, and gross profit margins decline year by year;

- Solution Customization: Masters 3D+AI algorithms, focuses on single scenarios, has a full order book but slow expansion;

- Platform Ecosystem: Develops its own algorithm platform, opens its SDK to small and medium-sized integrators, similar to an "industrial vision version of Android." Leading companies have already secured two rounds of financing, and their valuations are rapidly increasing.

In the next five years, platform ecosystem companies are expected to replicate the "cloud service" path—first collecting industry data through software subscriptions, then outputting high-margin standardized hardware, ultimately capturing most of the market profits.

3. Downstream: Customers Shift from "Buying Equipment at Single Points" to "Buying Capabilities for the Entire Line"

In the past, customers only considered "detection accuracy" and "manual replacement" in bidding. Now, tender documents include MTBF (Mean Time Between Failures), CT (Detection Cycle Time), False Call Rate, model iteration cycle, and data ownership in the contract. According to a survey by China Research & Consulting, in 2025, the top 20 power battery manufacturers will have as many as 72 performance indicators for visual inspection suppliers, three times the number in 2020. This means that "delivering a set of equipment" has upgraded to "delivering continuously upgraded inspection capabilities," and manufacturers without algorithm self-iteration and data closed loops will be rapidly marginalized.

III. Technology Trends: 2D to 3D is just the "entry ticket," AI generalization capability is the "moat"

1. 3D Vision Enters the "Cost Sweet Spot" The prices of line laser and structured light modules have dropped by more than half in the past three years, and domestic manufacturers have shortened the cost recovery period for introducing 3D vision into lithium battery top cover welding inspection to less than 9 months. China Research & Consulting predicts that by 2027, the penetration rate of 3D vision in industrial inspection will increase from the current approximately 20% to over 30%, with lithium batteries, automotive welding, and consumer electronics metal frames being the three core scenarios.

2. AI Algorithms Move from "Customization" to "Large-Scale Modeling" Industrial scenarios have many defect types and few samples, requiring traditional CNNs to be trained with large amounts of data. Starting in 2024, leading domestic vision vendors pruned and distilled large models such as SAM (Segment Anything) and "Shusheng Industrial" for defect segmentation, resulting in a 70% decrease in data annotation for new projects and a reduction in the deployment cycle from two months to two weeks. Zhongyan Puhua predicts that companies with the capability of "large industry models + rapid adaptation to small samples" will experience a surge in orders in 2026.

3. Cutting-edge technologies such as quantum imaging and event cameras are moving out of the laboratory. Quantum imaging has a natural advantage in low-light and smoky environments and has been included in the "challenge-based" list by the China Aerospace Science and Industry Corporation. Event cameras, with their high dynamic range and low latency, have completed proof-of-concept (POC) applications in high-speed collision detection and lithium battery stacking. Although industrialization will take time, the Zhongyan Puhua technology team warns that once the cost curve declines sharply, traditional 2D/3D solutions will face a "dimensional reduction attack." IV. Zhongyan Puhua's "15th Five-Year Plan" Outlook: Three Main Lines, Four Major Tracks, and Five Risks

1. Three Main Lines

- Domestic substitution is moving from "shallow waters" to "deep waters"—from cameras and lenses to AI algorithm platforms, comprehensively replacing foreign products;

- Business model is shifting from "selling hardware" to "selling capabilities"—subscription-based, pay-per-test, and cloud-edge collaboration;

- The focus of competition is upgrading from "single-point accuracy" to "complete production line data closed loop"—whoever can help customers improve yield by one percentage point will reap the excess profits.

2. Four High-Growth Sectors

- Power Batteries: Electrode slitting, sealing and welding, and post-weld module inspection; leading the industry in compound annual growth rate from 2025-2030;

- Photovoltaics: TOPCon, HJT, and perovskite multilayer film inspection, demanding higher spectral and higher resolution vision;

- Semiconductors: Advanced packaging RDL and μ-bump inspection; the shortage of domestically produced equipment creates opportunities for "bundled sales" of vision systems;

- Humanoid Robots: Six-dimensional force perception + visual fusion; exponential growth in demand for 3D vision modules; expected to become the fourth largest application market by 2030.

3. Five Potential Risks

- Geopolitics: High-end CMOS and GPUs are once again added to the sanctions list, and the progress of domestic substitution is slower than expected;

- Data Security: Cross-border transmission of industrial image data is restricted, and foreign cloud service providers are withdrawing from the Chinese industrial vision market;

- Talent Shortage: The shortage of compound talents who understand optics, AI, and industry processes continues to widen;

- Price War: Low-end 2D vision is highly homogenized, and some manufacturers are bidding at a loss to secure orders, leading to an overall decline in industry profit margins;

- Technology Route Shift: Disruptive technologies such as quantum imaging and event cameras are maturing ahead of schedule, resulting in asset devaluation of traditional solutions.

V. Action List for Different Types of Players

1. Upstream Core Component Manufacturers

- Seize the window of opportunity for dual certification of "automotive-grade + industrial" products to secure Tier 1 suppliers and system integrators in advance;

- Collaborate with domestic AI chip manufacturers on "hardware and software integration" optimization to achieve world-class power consumption and latency indicators;

- Extend downstream through industry funds and strategic investments to ensure early demand feedback. 2. System Integrators

- Upgrading from a "project-based" to a "platform-based" model – developing self-configurable and scalable algorithm platforms to reduce customization costs;

- Deepening focus on a high-growth sector (lithium battery, photovoltaic, or semiconductor), turning process know-how into data assets;

- Partnering with insurance companies and financial leasing companies to launch "visual inspection service packages," sharing profits based on yield improvement, reducing one-time CAPEX pressure for customers.

3. Terminal Manufacturers

- Incorporating visual inspection into the "factory infrastructure" budget, rather than the "equipment" budget, to enjoy the dual benefits of national large-scale investment funds and technological upgrading subsidies;

- Adding "data ownership" clauses to the bidding process to avoid future vendor lock-in;

- Establishing "joint laboratories" with universities and vision technology manufacturers to proactively develop next-generation quantum imaging and event camera technologies.

4. Investment Institutions

- Focus on teams possessing a three-in-one capability: large-scale model + industry data + hardware capabilities;

- Avoid "pseudo-growth" companies that only perform simple 2D integration, and be wary of sharp declines in gross margins due to price wars;

- Prioritize targets that have already entered the front-end testing of power batteries and semiconductors to enjoy the guaranteed order dividends over the next five years.

VI. Conclusion: Visual Inspection is the Key to "Self-Reliance and Controllability of High-End Equipment" Looking back over the past decade, China's manufacturing industry has captured global market share through low-cost, large-scale operations. Standing at the threshold of the "15th Five-Year Plan," we face even more stringent requirements for yield, efficiency, low carbon emissions, and supply chain security. As the "eyes of high-end equipment," the depth of localization of visual inspection directly impacts the competitiveness of strategic industries such as semiconductors, new energy, and automobiles. Zhongyan Puhua Industry Research Institute predicts that from 2025 to 2030, visual inspection will not only be a market worth hundreds of billions of yuan, but also a decisive battleground for whether China's high-end manufacturing can completely break free from the "bottleneck" problem.

Jinan Maotong, relying on its professional data research system, systematically collects, organizes, deeply mines, and accurately analyzes massive amounts of industry information, committed to providing customized data solutions and strategic decision support services for various clients. Through scientific analysis models and industry insight systems, we help our partners effectively control investment risks, optimize operating cost structures, discover potential business opportunities, and continuously enhance their market competitiveness.

For more cutting-edge industry insights and professional research results, please refer to the latest report released by Zhongyan Puhua Industry Research Institute, "2025-2030 China Visual Inspection Industry Competition Analysis and Development Prospect Forecast Report." This report, based on a global perspective and local practices, provides authoritative reference for corporate strategic planning.