We have three weeks left in the year, which means it’s time to stop guessing what “Industrial Internet” actually delivers and look at where the money is going. The industry is pivoting from the speed-and-scale growth of 2019–2024 to a new cycle defined by efficiency, profitability, and survival.

At the 2025 Ebrun Industrial Internet Annual Conference in Beijing, nearly a thousand practitioners gathered to discuss this rotation. Zheng Min, Chairman of Ebrun Power, outlined three strategic keys for this new era: Industrial AI, Deep Value Chains, and Going Global. These aren’t just buzzwords; they are the mechanisms driving high efficiency and fostering digital supply chain platforms for large-scale organizational ecosystem collaboration.

In practice, if your platform doesn’t solve latency or cost issues in these verticals, you’re a lab demo, not an operator. I think data silos are still the biggest blocker to AI; fix the plumbing before buying the model.

Song Zhiping of the Association for Listed Companies noted that while past growth prioritized speed, slowing markets now demand quality and efficiency. Pan Yong of Guolian Shares added that the industrial internet’s role is helping SMEs and traditional industries navigate this future. “When others succeed, you make money; when others fail, you survive,” he said, emphasizing the need for clear positioning. Liu Taoran of Lange Group stressed persisting in doing things that are “one meter wide but a hundred meters deep.”

During the gala, Ebrun Think Tank released the 2025 Industrial Internet Development Report, alongside the “Value Creation and Cooperation” closed-door meeting and the “Night of Thousand Peaks” (Qianfeng Ye). This article breaks down those insights through the lens of the three strategic directions.

Industrial AI: The Engine of the New Cycle

In bulk commodity trading, understanding market trends is essential for profit. Behind sharp price fluctuations lie hidden opportunities and risks. Liu Taoran stated that using AI tools to predict futures and spot prices offers calculation speed and capacity far exceeding human analysts. This replaces previous methods of guessing market trends based on experience or gambling on the market.

This isn’t an isolated case. 2025 financial reports from many listed industrial internet companies revealed AI penetration into R&D, production, logistics, sales, and decision-support scenarios. The shift is clearly toward “pragmatism,” moving away from parameter-chasing model competitions to deep cultivation and large-scale application within specific scenarios.

Operationally, stop chasing parameters; deploy models that reduce operational variance in your core workflows. In practice, rOI on AI is only visible when it replaces expensive human guesswork with deterministic speed.

Vertical industrial platforms are the main battleground for AI implementation. From an application perspective, enterprises must first solidify their data foundations and break down information silos to effectively utilize AI. Vertical industrial platforms possess natural advantages in driving large-scale AI applications due to their high-quality data covering the entire chain, deep understanding of the industry, and market base linking massive numbers of enterprises.

“The industrial internet itself is a platform that connects the entire industrial chain online and accumulates key industrial scenarios and data,” said Wang Ke, Senior Vice President of Nongxin Digital Intelligence. “With these fou

I read through the latest insights on industrial AI, and frankly, the shift from hype to operational reality is what matters. The industry is moving past “appropriationism” toward a pragmatic architecture: large models for navigation, small models for execution. This isn’t just theory; it’s about reducing latency and cost in production environments.

Wang Ke, Senior Vice President of Nongxin Digital Intelligence

The core strategy emerging is “Large Models for Navigation, Small Models for Execution.” Ding Deming, Vice President and General Manager of Strategic & Business Development at JD Industrial, outlines that large models handle expert brains and underlying knowledge bases, while domain-specific small models manage execution. Agents then perform efficient collaborative scheduling, creating a path that is professional, reliable, economical, secure, and feasible.

Ding Deming, Vice President and General Manager of Strategic & Business Development at JD Industrial

The data supports this shift toward efficiency. In 2025, agile enterprises are building dedicated AI models based on industry knowledge to solve specific pain points. Over the past three years, domestic “≤10B parameter” small models have become the fastest-growing segment in the large model landscape. Their release share rose from approximately 23% in 2023 to over 56% in 2025.

I think small models reduce inference costs and latency significantly compared to monolithic giants. This is shippable today, not a lab demo.

Specific results are already visible in the field. Zhenkunhang’s self-developed industrial AI large model, “Xingjia Linglong,” achieved a product selection accuracy rate of 98%. By comparison, general large models hit only 66%, and traditional models sat at just 30%. Nongxin Digital Intelligence now serves tens of thousands of pig farms. Supported by AI, smart hardware, and real-time data, their algorithms effectively manage farming operations.

Wang Ke explained that the feeding model focuses on growth speed after feed consumption. It is not a general model but allows enterprises to build proprietary systems. Previously, an inventory count at a pig farm took 10–15 days; with AI and smart cameras, it now takes only two to three minutes, enabling financial innovations. A plant manager could previously oversee 5,000 or 50,000 pigs; now, they can manage 500,000, 5 million, or even 50 million pigs using AI assistance for core decision-making.

Operationally, automating inventory counts from days to minutes eliminates a major on-call bottleneck. This directly reduces operational overhead and error rates.

Agents are now addressing the “last mile” problem of integrating large models with industry. They help pass on employee experience, make implicit knowledge explicit, and reduce repetitive labor. At JD Industrial, employees use AI for efficiency while departments focus on compliance and risk control. At Yunzhonghe, AI Agents are applied in member marketing, digital benefits, corporate centralized procurement, and private domain e-commerce.

Dong Hui, Chairman of Yunzhonghe, argues that top executives must shift from the tactical view of “how to improve efficiency” to the strategic view of “how the enterprise moves toward transformation.” AI Agents are currently being used at strategic and management levels. Zhao Fengwei, Founder and CEO of Duan Dian Technology, stated that AI consultants provide insights into global business operations and decision-support capabilities, offering risk decisions and recommendations for specific metrics and business progress.

In practice, strategic oversight via agents doesn’t replace the need for robust monitoring pipelines. We still need to own the data quality feeding these systems.

SMEs are leveraging platform AI tools to achieve capability leaps, adopting strategies of “following the platform, leading the platform, and transcending the platform.” In 2025, many platforms made industrial AI a key strategic direction, opening product capabilities to upstream and downstream enterprises. For instance, 1688 launched an AI version of its App, an AI version of its Chengxintong (TrustPass) service, and AI agents this year. The platform encourages merchants to open stores on 1688 and do business with AI. According to data disclosed by the platform, in the first month of operation for the Chengxintong AI version,

I read through the latest insights on the industrial internet’s new cycle, and what stood out wasn’t just the hype—it was the operational reality. Merchant GMV grew by 73%, inquiries increased by 15%, buyer numbers rose by 20%, and repurchase rates improved by 31%. These aren’t lab metrics; they’re signals that AI is finally moving from pilot to production in supply chains.

Opening up AI capabilities on platforms not only helps merchants reduce costs, improve efficiency, and achieve data-driven operations but also supplements capabilities and lowers barriers. This brings capability upgrades to SMEs and promotes the rise of new organizational forms such as “one-person companies” and “smart production lines.” (Reference: Ebrun’s Zheng Min: What Does AI Equality Mean for SMEs?)

According to cases in the 2025 Industrial Internet Development Report, Shu Kai, a second-generation successor to a Yiwu towel factory, used AI design to quickly produce circular beach towels suitable for the Tunisian market, tripling design efficiency. The Dongguan Niuding Titanium Jewelry Factory activated the Chengxintong AI version and completed over 40,000 orders in just 15 days, rising to the top of the platform’s nose ring category.

“Originally, you had to conduct market research, R&D, production, and distribution sales from scratch. Now, you can hand over low ROI tasks to platforms with stronger capabilities and differentiation, focusing your efforts on accumulating core competencies,” said Steven, General Manager of 1688 Industrial Intelligence Alliance. “What ultimately forms may be the true form of industrial interconnection.”

Steven, General Manager of 1688 Industrial Intelligence Alliance

Deep Value Chains Will Not Stop at Transactions

China is the country with the most complete industrial categories in the world, with total industrial supply chain costs reaching 115.19 trillion yuan in 2024. Ding Deming, Vice President and General Manager of Strategic & Business Development at JD Industrial, believes that while each industry may be “one meter wide,” connecting deep nodes in the supply chain is equivalent to extending it “100 meters deep.” This goes beyond transactions to release the space for full-link collaborative effects.

“To date, China’s industrial sector has not completed its transformation,” Ding said. “This requires a long process, possibly five or ten years.” Entering a new stage, AI acceleration and penetration are driving changes in supply and demand structures. It is not just about reducing costs and improving efficiency but also reconstructing supply chain structures to form new value spaces. This makes the deep value chain the second key to unlocking the new cycle.

Currently, explorations and practices at the industrial level have blossomed in multiple directions. Ebrun Power & Ebrun Think Tank has long conducted tracking research on the industrial internet. Based on the thinking and practices of advanced enterprises, we propose a three-dimensional structure for deep value chains:

• Vertical Integration: Breaking down data silos across “production, supply, and sales” to achieve demand-driven production and agile response;
• Horizontal Integration: Integrating technical, financial, park, and other supply chain supporting resources to build a stable and efficient service ecosystem;
• Breaking Through Standards, R&D, Brands, and Data Assets: Reconstructing value upwards to thicken profit margins.

This three-dimensional structure promotes the transformation of industrial chains from linear “cost competition” to networked ecosystems of “value symbiosis,” serving as the core practical framework for releasing the vast potential for value creation in the industrial internet. “A forest cannot consist of only one tree; it needs other trees, flowers, and grass—that is an ecosystem,” Song Zhiping also noted. “(The industrial internet) involves both competition and cooperation, actively maintaining industry interests.”

Song Zhiping, Renowned Management Expert and Chairman of the Association for Listed Companies in China

Vertical Integrated Industrial Chain Deep Collaboration. Some leading enterprises use AI to connect internal links such as “production, supply, and sales,” building a supply chain system driven by demand that allocates upstream production and resources.

I think siloed data kills agility; vertical integration is non-negotiable for real-time ops. Operationally, aI design efficiency gains are only valuable if they ship without breaking the build. In practice, cost reduction means nothing if latency spikes during peak order volumes.

I read through the latest industry reports and case studies to see what’s actually moving the needle in industrial MLOps this week. The shift from “single-point efficiency” to “global optimization” is no longer just marketing fluff; it’s a hard requirement for survival as costs climb.

According to the 2025 Industrial Internet Development Report, Guolian Shares empowered a titanium chloride production enterprise through its “Deep Supply Chain + Digital Factory” model, increasing capacity to a maximum of 11,000 tons per month. The operating cost per ton decreased by approximately 700 yuan, and procurement costs dropped by about 1,500 yuan per ton.

I think real latency reduction in supply chain logic beats theoretical AI promises every time.

Wang Fei, Chairman and General Manager of Shaanxi Coal Material Mall Company, believes that the fastest implementation of AI-driven deep industrial chain collaboration is in procurement bidding. “Procurement bidding involves publishing documents, opening and evaluating bids, and expert review. It fully relies on AI for reviewing documents, qualifications, and other technical and commercial terms,” she said. The collaborative effect kept the procurement price at a 6.5%-6.8% discount compared to market prices, shortened the procurement cycle from half a month to two days, and increased per-capita procurement volume from 6.5 million yuan to 28 million yuan.

Operationally, automating bid evaluation cuts cycle time by 90%, which is an immediate ROI win for ops teams.

Horizontal Integrated Supply Chain Digital-Physical Support

Some leading enterprises use digital platforms as hubs to integrate scattered, specialized external resources such as logistics, warehousing, finance, technology, and industrial parks. This not only provides a stable, efficient, and low-cost support system for business operations but also builds ecological collaborative capabilities for data connectivity and process reconstruction.

For example, Jiwei IoT integrates idle social and industry resources to build a full-lifecycle service ecosystem for bulk commodities, linking over 110,000 upstream and downstream enterprises. Its annual business volume exceeds 120 billion yuan, reducing logistics costs for manufacturing enterprises by over 8% annually and increasing income for warehousing and processing service enterprises by more than 15%.

Similarly, Shaanxi Coal Material Mall Company integrates other e-commerce platforms as needed for MRO (Maintenance, Repair, and Operations) categories. Cui Wei, Chairman of Zhongtie Wumao Group’s Jiwu Technology, believes that cross-organizational boundary industrial chain collaboration may be one direction for deepening value chains in the future, but it also involves issues with coding standardization.

“Traditional methods rely on API integration, but many specific problems remain,” Cui Wei said. The current pain point in the industrial goods sector is precisely the lack of standardized coding; therefore, this shortcoming should be prioritized for improvement. Wang Bo, Executive General Manager of CCCG E-commerce Company, also believes that the application of AI technology must be based on solid underlying data management, such as coding standardization.

In practice, without standardized coding, your AI models are just guessing on unstructured data.

Industrial Product Branding Breakthroughs

Beyond internal and external synergy, some industrial internet enterprises are also driving industrial products from homogeneous competition characterized by “non-standard low prices” toward brand development based on “standardized high quality.” Chen Long, Chairman of Zhenkunhang Industrial Supermarket, believes that with the popularization and promotion of AI, barriers in the industrial product supply chain will rise. Enterprises must build core competitiveness through deepening AI-driven product innovation, quality improvement, and personalized services.

Some industrial product companies have already realized the importance of branding and are gradually advancing domestic substitution in key areas. For example, Zhenkunhang has created a matrix of specialized sub-brand products in fields such as personal protective equipment, tools, general consumables, administrative supplies, general equipment, and smart warehousing. The number of clients served by its private-label brands has reached tens of thousands.

In fact, in deep value chain practices, data that has undergone systematic collection, governance, modeling, and application can be directly converted into corporate assets through securitization. In September 2025, Shaanxi Construction Logistics issued the “Huaxin-Xinxin Data Asset Phase I Asset-Backed Special Plan” on the Shenzhen Stock Exchange, with an initial issuance scale of 133.7 millio

I read the latest analysis on China’s industrial internet expansion, and the numbers are stark. In 2024, total manufacturing exports hit $3.26 trillion, edging out Germany, the US, and South Korea combined ($3.25 trillion). But here is the catch: in many sectors, Chinese firms still lack decision-making power. They are volume leaders, not value architects.

Industrial Global Expansion: Capturing Dividends from Supply Chain and Brand Globalization

While China dominates volume, Western players like Grainger, Fastenal, and Würth sit atop the MRO (Maintenance, Repair, and Operations) value chain. Fastenal’s 2024 results prove this model works: $7.347 billion in revenue (+5.2% YoY), $1.155 billion net profit (+6.3%), and a razor-sharp 15.72% net margin. That is the kind of operational efficiency we chase in platform engineering—high margin, low friction.

I think high-margin MRO models prove that service layers beat pure hardware volume every time.

Geopolitical headwinds are forcing Chinese firms to rethink their global strategy. Chen Long, Chairman of Zhenkunhang Industrial Supermarket, notes the danger: “If you only provide supporting components for others, you are merely a supplier, subject to price suppression and low gross margins.” The goal isn’t just export; it’s building industrial giants that control their own destiny.

The era of blind price wars is ending. Entrepreneurs are realizing that “involution” destroys value. Instead, companies must climb the smile curve—organizing upstream resources and distributing downstream interests. Liu Bo, Founder of Pandding International Group, argues that globalization is no longer just moving goods; it’s an ecological expansion of capacity, brand, and technology.

Using Warehouses to Drive Chains: Overseas Warehouses Upgrade Supply Chain Hubs.

The “warehouse-driven chain” model treats overseas warehouses as physical pivots and data hubs. It integrates forward head-haul transport with backward local delivery, installation, and after-sales support. This creates a closed loop of “transportation, warehousing, distribution, and sales.”

Pandding International exemplifies this with 70+ overseas warehouses totaling 600,000 square meters. They offer more than freight forwarding; they handle customs compliance, showrooms, and after-sales for SMEs. For smart home manufacturers, this means less operational drag and faster time-to-market in foreign markets.

Operationally, centralized logistics data reduces latency in the supply chain, turning warehouses into active nodes rather than static storage.

“We have also begun to focus on fields such as smart products, high-end medical equipment, outdoor sports, industrial equipment, and food since last year,” Liu Bo says. He highlights that heavy machinery is accelerating its global push via the Belt and Road Initiative, aiming to become a second growth curve for Chinese exports.

Using Products to Drive Chains: Helping Chinese Industrial Clusters Land Globally.

China holds 35% of global manufacturing output, backed by 300 industrial clusters and over 6 million enterprises. Yet, SMEs struggle to go solo. The solution is “going global as a group,” building an industrial community system to share resources and risk. This collective approach allows smaller players to access the infrastructure they couldn’t afford individually.

In practice, shared infrastructure lowers the barrier to entry for SMEs, but requires robust API integration to avoid siloed data.

Unlocking the New Cycle of Industrial Internet: Key Insights

Guolian Shares has carved out a niche in vertical industries defined by concentrated upstream competition and fragmented downstream markets. By integrating logistics, finance, digital transformation, and certification services across seven overseas operation centers and a network of warehouses, they are pushing Chinese industrial clusters into global markets. This model relies on breakthroughs with core single products to drive category synergy, offering a replicable path for traditional industries going global.

I think centralized procurement platforms reduce latency in supplier matching but introduce single points of failure if the digital backbone goes down.

State-Owned Enterprises Enter via Digital Procurement

Central and state-owned enterprises are leveraging supply chain integration, large-scale procurement advantages, and compliance systems to enter the cross-border battlefield. Unlike traditional trading firms fighting alone, these platforms use centralized procurement to cut costs and increase efficiency. Simultaneously, they employ digital means to ensure process traceability and risk control, effectively exporting China’s supply chain capabilities.

A prime example is “Runhuicai,” operated by China Resources. Originally built for internal operations, it has evolved into an overseas procurement coordination platform for the Greater Bay Area, providing standardized cross-border services. Ran Peng, General Manager of China Resources Shouzheng, notes: “It provides excellent tools for central and state-owned enterprises going global in terms of processes, platforms, and compliance, attracting companies to post their demands on the platform.” He adds that it builds a transparent matchmaking system that attracts high-quality suppliers seeking business opportunities.

Operationally, traceability is critical for audit trails, but real-time data synchronization across borders often becomes the bottleneck during peak load events.

Industrial Internet Thinking Builds Global Brands

Going global now requires an ecosystem upgrade rather than just marketing and channel expansion. Some brands are using data intelligence and user co-creation to build dynamic, resilient value networks through ecological collaboration that competitors struggle to imitate quickly.

Leqi Innovation’s SmallRig brand exemplifies this approach. Specializing in imaging accessories, it leverages supply chain capabilities to create an agile model characterized by “small batches, frequent deliveries, diverse categories, high frequency, and semi-customization.” The brand has established a “user-manufacturer co-creation” development cycle, launching over 500 innovative products annually (roughly 1.6 per day) with a basic closed-loop development time of just 21 days.

Leqi Innovation has also achieved industry synergy through digital process integration. Gao Haiyan, Co-founder of Leqi Innovation SmallRig and Chairman of the Leqi SmallRig Imaging Development Fund, explains: “Upstream supply chains, innovation hubs, and startups can all initiate innovations. Through digital process integration, they choose co-creation methods on an external digital co-creation platform, matching business models and profit distribution.”

In practice, a 21-day product cycle is impressive, but it demands robust CI/CD pipelines for hardware firmware to avoid shipping bugs at scale.

Gao Haiyan outlines a three-stage evolution for brands going global: first as cross-border e-commerce entities selling via third-party platforms; second as internet companies focusing on intensive scale operations and user value creation; and finally moving toward the industrial internet, evolving from a value ecosystem to an industrial ecosystem.

He argues that Leqi Innovation’s brand loyalty stems from ecological scale, value delivery, and supply levels working together, rather than traffic investment. “Moving from products to categories, scenarios, and ecosystems inevitably requires industrial internet methods,” Gao Haiyan states. “It is both an active strategy and a passive choice.”

Summary

People often overestimate changes in one year but underestimate changes over ten years. As old and new cycles alternate and the economic environment shifts, many feel lost. However, if we extend the timeline, helping small and medium-sized enterprises and traditional industries move toward the future should be the long-term narrative.

Unlocking the New Cycle of Industrial Internet: Key Insights

The shift from consumer internet dominance to industrial application is no longer theoretical; it’s an operational reality. Song Zhiping argues that China’s 30% share of global manufacturing output, combined with early infrastructure advantages in cross-border e-commerce and logistics, grants absolute discourse power in this new era. The strategy is clear: leverage existing scale to lead the industrial internet transition.

I think scale in manufacturing doesn’t automatically translate to low-latency industrial control. We need deterministic networking, not just e-commerce throughput.

On December 3, the 2025 Ebrun Industrial Internet Annual Conference debuted the 2025 New Cycle Navigation Case Collection. This isn’t a lab demo; it’s a curated list of twelve enterprises selected for outstanding practical achievements: Guolian Shares, Zhenkunhang, DuanDian Technology, Qixin Group, OFS, Xinfangsheng, Jiepei, Lange Group, Yunhan Chip City, Sanheng Technology, Xingpingtai Group, and Shaanxi Construction Logistics.

Operationally, these are the vendors shipping real integrations today, not just pitch decks. Prioritize their stack for immediate reliability gains.

Appendix: List of 2025 Qianfeng Award Winning Enterprises

In practice, vendor lock-in is a risk, but proven interoperability reduces on-call pain significantly.