As the Chinese New Year approached in 2025, the author had the opportunity to engage with MrXiao, who has been running a machine embroidery business, and MrFang, the head of Guangzhou Yijiu Information Technology Co., Ltd., which specializes in AI researchThey established a WeChat group to explore the integration of AI design with embroidery techniquesHowever, due to the hustle of the festival, they weren't able to meet in-depth until the ninth day of the lunar new year, when the author accompanied MrXiao to MrFang’s office to discuss how AI could empower the embroidery industry.

In their discussions, MrFang eagerly sought to understand the embroidery industry's dynamics and requirementsLeveraging advanced AI models, he continuously created stunning embroidery designs that showcased a variety of styles such as European, Indian, Arab, Japanese, and Chinese aestheticsEventually, he even generated a series of fashion pieces inspired by these embroidered artworks, which were nothing less than exquisite.

This left MrXiao in deep reflection, acknowledging that the integration of AI into manufacturing wasn't just a theoretical notion—it had become a tangible realityThirty years ago, he founded Guangdong Haidi Jun Embroidery Co., Ltd., and over the years, his company has risen to become an invisible champion within this sector, manufacturing products commonly used in fashion, bedding, and home wear.

Reflecting on traditional Chinese craftsmanship, MrXiao felt that it often emphasized technique but lacked a refined sense of artistic aesthetics, imagination, and contemporary expressionAdditionally, traditional crafts were struggling against industrialization and issues of inheritanceHe had long aimed to fuse traditional embroidery techniques with modern digital production technologies to create contemporary embroidered pieces that were more three-dimensional, uniform, vivid, and artistically expressive.

In recent years, Mr

Advertisements

Xiao had been contemplating the transformation of his business and originally intended to combine machine embroidery with artistic creation to elevate his company from manufacturing to artistic innovationHowever, after engaging with AI design, he found his thought process transformedHe recognized that AI could facilitate comprehensive digital transformation and upgrades for businesses, prompting a shift in business models and addressing inheritance issues within companies.

The Real Integration of AI with Manufacturing

In the two years prior, discussions surrounding ChatGPT and its impact on humanity had intensified, with industries pondering the future of AI integration in various sectorsThe Guangdong SME Development Promotion Association organized multiple seminars and training sessions, aiming to unite professional teams to develop fusion scenarios between AI and manufacturingHowever, the results were broadly seen as unsatisfactory, with many feeling that the advancements were still shrouded in uncertainty, merely providing a fresh perspective.

The excitement surrounding the emergence of DeepSeek among the Chinese populace stems from the rapid rise of large-scale AI models that have resolved the question of usability and are now poised to tackle the practicality challenge, which has become a real demand within the industry.

After attending several AI workshops organized by the promotion association, MrFang noted that although Guangdong might have missed out on DeepSeek, the true opportunity lies in the deep fusion of DeepSeek with manufacturing—specifically, the widespread application of large-scale models in manufacturing contexts throughout Guangdong.

According to the chief advisor of the promotion association, Mr

Advertisements

Hao, should DeepSeek achieve a 30% penetration rate within Guangdong's manufacturing by 2027, it could generate an economic value increment of 300–450 billion yuanMrHao argued that if Guangdong missed the chance for comprehensive application of DeepSeek in manufacturing, it would not only lose potential economic growth but also relinquish technological sovereignty in the manufacturing sector, leading to a fragmented chain in computing power, algorithms, and data.

The first concern is the hollowing-out of computing infrastructure, while intelligent computing centers in Zhejiang and Shanghai may monopolize model training nodes.

Secondly, there’s a risk of data asset outflow; industrial data processed via external platforms weakens local data sovereignty.

Thirdly, the risk of widening efficiency gaps in manufacturing persistsDespite Guangdong currently leading in manufacturing scale with a projected output value of 4.2 trillion yuan in 2024, its total factor productivity (TFP) only stands at 68% of that in the United StatesIf DeepSeek's application is not scaled, industries in Jiangsu and Zhejiang through AI empowerment will likely widen efficiency disparities.

Fourth is the potential reversal of talent siphoning effectsCurrently, Shenzhen has an AI talent density of 12.7 per ten thousand citizens, less than Hangzhou's 15.3. Should DeepSeek's application lag, talent might migrate from Guangzhou and Shenzhen to the Yangtze River Delta region, creating a ‘migration of innovators’.

It can be observed that Guangdong, with over 7 million SMEs spanning more than 30 sub-sectors, including electronics, home appliances, and clothing, has a highly dispersed industrial structureThis results in an ‘inverted tail’ distribution of AI application scenarios—each sub-sector appearing minor in demand, yet collectively presenting an enormous value gap.

The production scenarios of SMEs are often characterized by fragmentation and high complexity

Advertisements

For instance, temperature control in kilns within the ceramic industry in Foshan, electronic manufacturing in Dongguan, and the flexible assembly lines in Zhongshan’s lighting industry require that AI transformations must engage deeply with industrial details to elevate general algorithms to specialized applicationsThis ensures a genuine fusion of digital and industrial engineering.

SMEs as Key Players in the AI Transformation of Industries

AI is poised to spark a new wave of industrial revolution, and the AI layout for SMEs is central to this battleground.

Firstly, SMEs form an essential component of the industrial ecosystemAs AI technology redefines manufacturing, the competitiveness of SMEs relies not merely on isolated technological breakthroughs, but rather on systemic optimization within the industrial ecosystemThis optimization is driven by enhanced cost-management capabilities and improved efficiency mechanisms, reshaping the regional and even national industrial standing.

The case of Siemens' Chengdu digital factory validates this logic: its global competitiveness and status as a ‘lighthouse factory’ stems not solely from its technical prowess, but also from the innovative iterations of China's supply chain networks and its capacity for cost control, heavily reliant on deep cooperation with local supply chains.

The Chengdu factory produces 2,300 types of products daily, supported by a supply chain comprising precision components from the Yangtze River Delta, electronic components from the Pearl River Delta, and logistical networks throughout the Chengdu-Chongqing areaThis ecosystem allows Siemens to rapidly respond to order changes, compressing delivery times to 60% of industry averages.

Moreover, an SME in Dongguan supplying connectors to Siemens enhanced its capacity variance tolerance from ±15% to ±35% through AI-driven dynamic scheduling systems, becoming a pivotal support pillar in Siemens' supply chain flexibility.

Interestingly, while the unit labor costs at Siemens' Chengdu facility stand at just one-quarter of those at its Amberg plant in Germany, the defect rates for products are even lower

This contrast roots in the speed of supply chain responsiveness and the efficiency of technology diffusionThe average delivery cycle for suppliers at Siemens Chengdu is 48 hours compared to 120 hours in GermanyChinese SMEs can adapt to AI processes within three months, while their German counterparts often take over a year.

The shift in competition within Chinese manufacturing has pivoted from ‘low-cost production’ to ‘high-flexibility ecosystems’.

At the same time, the collaborative effects of technology are increasingly emergingIn the process of digital transformation chains among enterprises in Guangdong, leading companies such as Midea share AI quality inspection algorithms and energy optimization models with SMEs, while the latter provide production line data to refine algorithm iterationsThis mutual empowerment has seen overall defect rates within Midea’s supply chain drop by 40%, and the average customer acquisition costs for SMEs decrease by 28%. As per data from the Guangdong Ministry of Industry and Information Technology, SMEs participating in chain transformation show AI application efficiency rates three to five times higher than those undergoing isolated transformations.

The industrial ecosystem formed by SMEs is pivotal to the overall industry’s competitive strengthIn a significant way, the efficiency of SMEs dictates the efficiency of the entire industry.

Secondly, the cost control within Chinese manufacturing is evolving from “scale dividends” to “intelligent dividends.” The cost advantage of Chinese supply chains has shifted from ‘labor price differences’ to ‘cost reduction through intelligent collaboration’. For instance, an industry cluster in a particular region has leveraged an AI material scheduling platform to boost the regional inventory turnover from eight times a year to fourteen, effectively liberating 12 billion yuan in floating capital.

This cost optimization cannot be achieved by individual enterprises; it requires a collaborative ecosystem characterized by data-sharing and algorithm collaboration

In Zhongshan's lighting industry cluster, 30 SMEs shared an AI design platform, reducing the mold development cost from 120,000 yuan to 30,000 yuan while compressing the product iteration cycle from six months to just 45 days.

The unique characteristic of AI technology lies in its near-zero marginal costsThe initial investment to develop an AI kiln temperature control model for the ceramic production in Foshan was 2 million yuanWhen scaled across 100 enterprises, the allocation cost per enterprise is just 20,000 yuan, yet can eliminate energy consumption by 15% to 20%. This characteristic makes SME clusters the biggest beneficiaries of the AI dividends.

Furthermore, improvements in manufacturing efficiency mechanisms are shifting from “experience-driven” models to “data-driven” paradigmsOne enterprise in Wenzhou, focusing on smart manufacturing for paper cups, achieved digital management that accurately enhanced materials' inventory management, shortened delivery cycles by 20%, and eliminated over 50% of non-value-adding activities, resulting in a 10% reduction in management personnel mainly from mid-level management.

The Yiwu small commodities market in Zhejiang, through an AI design platform, boasts over 100,000 new SKUs every day, thanks to collaborative innovation across 3,000 SMEsThe design data from each enterprise enters a shared pool in real-time, powered by AI for trend forecasting and optimized combinations, leading to a threefold increase in blockbuster product output rate across the entire ecosystem—an innovation efficiency unattainable by individual enterprises.

In the competition between the Yangtze River Delta and the Pearl River Delta, differentiations emerge among Jiangsu’s chemicals, Zhejiang’s light industry, and Guangdong’s electronic manufacturing ecosystems

Once any region establishes a rapid uptake of AI technology among its SMEs, it can create a speed advantage, enhancing regional competitivenessFor example, the smartphone industry chain in Dongguan has utilized AI transformations among 500 medium-sized EMS enterprises to shrink the new product trial production cycle from 45 days to just sevenThis rapid conversion gives brands like Huawei, OPPO, and others the confidence to compete with Apple’s launch cadence.

The optimization of ecosystems within SMEs will become the ultimate battleground in the AI eraThe competitive edge of China's industries fundamentally hinges on the health of its manufacturing ecosystems, where SMEs do not merely play the role of followers behind industry giants; rather, they serve as the source of ecological vitalityIn the future, regional competition will be a function of the "depth of AI technology × density of ecological collaboration," where technological depth determines single-point heights (such as algorithm accuracy and computational scale), while ecological density influences application breadth (including data mobility and technology diffusion speed).

For Guangdong and China's manufacturing sector to maintain their edge, the AI layout for SMEs must be seen as a core facet of ecological optimizationThis can be achieved by lowering technological barriers through chain transformation and constructing a new type of industrial relationship characterized by data sharing, standard interoperability, and mutual benefitsOnly thus can the Chinese supply chain evolve into a smart ecosystem capable of exercising defining power on the global stage, rather than merely adapting.

Challenges in Enabling AI for SMEs

In discussing with MrFang, MrXiao was first drawn to considerations regarding costs

Although MrFang argued that with DeepSeek's introduction, local deployment costs for businesses would significantly decrease, MrXiao still harbored numerous concerns, such as the rapid progression of AI developmentCould DeepSeek maintain its leading edge, and how might businesses effectively keep pace? Furthermore, with the integration of AI, how should companies restructure internally to accommodate AI-driven management and establish new strategies?

Indeed, at a discussion session organized by the author, many enterprises expressed a desire to embrace AI yet found themselves trapped in an anxious state of not knowing where to start.

While the trend towards digitization and intelligence across industries is an inevitable wave, SMEs predominantly hail from traditional business models, facing resistance rooted in their structural, cultural, or behavioral traits that impede technological transformationsThis resistance is not born from the technology itself but rather embedded in the deep operational logic within organizations.

The first challenge arises from organizational dynamicsDigital and intelligent transformations must be championed by leadership, yet the execution rests with teamsMany management layers view these shifts as burdensome endeavors fraught with complexitiesSenior executives often assume a mindset of maintaining the status quo; for many, simply collecting a paycheck equates to doing less work.

The promotion association's push for businesses to engage in cost-reduction projects has faced similar hurdlesHigh-level executive resistance often stems from an ingrained perception that success without disruption is impossible, compounded by a reluctance to integrate external inputs into their "territory." Advancing digitization means that the efficiency of a company accelerates; data-driven decisions diminish the role of intermediary decision-makers; packaging management knowledge into systems can diminish the significance of managers, discouraging their involvement in digital transformation efforts.

A company leader reflected on a decade-old experience with information technology transformations, describing it as turbulent but noting the faith that informed that choice

Without a relentless belief in the efficacy of informatization, enterprises cannot win in the battles of information, digital, or intelligence.

The crux of organizational inertia lies in the incompatibility between "human cognitive operating systems" and the algorithmic logic of AIFor firms to break through these limitations, they must simultaneously advance cognitive upgrades, skills re-engineering, process reconstruction, and cultural shifts to integrate AI into the very DNA of the organizationOnly then can they avoid attempting to drive intelligent vehicles using the outdated frameworks of the industrial era.

Secondly, there are issues regarding perception and recognitionA widespread belief exists that AI within businesses operates as a "cost center" rather than as a "strategic investment," leading to a preference for holding onto traditional production methodsSome argue that human experience outweighs algorithms, inherently disregarding processes that could significantly improve the quality of their productsAs competitors leverage AI to enhance product yield by 12%, these companies may see their market share decline.

Executives often lack an understanding of AI's long-term value, overly reliant on historical success paths like scale expansion and low-cost competition, thereby creating a disconnect between strategic decision-making and technological innovation.

Thirdly, traditional management systems bind innovationConventional corporate processes aim primarily to mitigate risks, while AI applications require agile responses and rapid iterations—culminating in a fundamental clash between these two frameworks.

Under traditional hierarchical structures, performance indices across departments can become siloed—for instance, the production department is assessed on yield while procurement focuses on cost—hampering the emergence of data-sharing incentivization mechanisms

Moreover, performance evaluations still center on traditional metrics like output and cost, leading employees to lack rewards for optimizing processes through AI.

This segmented performance evaluation often stymies new technological flexibility, as linear management workflows cannot accommodate the real-time decision-making demands of AI.

With each department managing portfolios in isolation, discrepancies in data standards arise, resulting in AI models failing due to data unavailabilityAn instance arose where one firm faced challenges during AI system development due to the production team's refusal to share real-time data, while the procurement department's Excel formats conflicted with data from MES systems, leading to a system prediction accuracy of less than 50%.

Fourth is the dichotomy between AI service provisioning and talent challenges within enterprisesGenerally, the dissemination of new technologies relies on specialized teams providing dedicated servicesYet these teams typically channel their energies into dominant industry players, failing to provide timely solutions for SMEs, often leading to missed opportunities.

Because many small startups develop vertical solutions without fully grasping industry-specific operational contexts, these offerings can struggle to find effective integrations into actual workflowsBuilding in-house AI teams also presents hurdles, with many SMEs grappling with a lack of AI talent and associated costs.

Finally, there's the issue of technical adaptabilityMost AI solutions are crafted with larger enterprises in mind, creating difficulties in meeting the fragmented demands of SMEsNotably, a survey conducted by the Guangdong Ministry of Industry and Information Technology indicated that 72% of SMEs perceived available AI products as having "excessive functionalities," with only 18% of features being actively utilized.

AI model training necessitates profound craft knowledge, yet many SMEs lack standing capabilities for data labeling and model fine-tuning

Current market solutions primarily emphasize intersections of AI with human resources, market marketing, and advertising designsIt is challenging for digital teams to fully grasp and integrate all modules of business managementEven when holding operational analysis meetings, businesses often resort to manually analyzing numbers, significantly affecting efficiencyIf internal data is incomplete or discontinuous, this would further hinder the enterprise’s AI potential.

Despite the numerous challenges they face, the author believes that various approaches can tackle these obstacles to strengthen the core battleground of industrial AI transformation.

First, specialized training programs should be initiatedFor instance, digital training for CEOs can incorporate case studies from within the industry to reshape decision-makers’ understandingsThis could include training modules based on AI knowledge for human resources, cost-reduction innovations, and corporate marketing; furthermore, a "Digital Craftsman" certification system should be implemented to incorporate AI competencies into technician title assessments.

Second, internal enterprise development must be prioritizedCorporations should regard digital and AI advancements as strategic directions of growth, establishing an organizational culture suited for AI progress by transforming from a "control-oriented" structure into an "agile ecosystem". Creating a “Digital Decision-Making Committee” within companies could also empower frontline workers to make micro-decisions based on AI insightsBesides, linking AI proficiency certifications with performance bonuses could create incentives for skill acquisition, ensuring knowledge engineers connect AI models with process expertise.

Third, nurturing AI services intermediaries, particularly focused on vertical fields such as furniture, garments, home appliances, lighting, and environmental protection, is equally crucial

Specialized platform service providers—integrating industrial design, innovative research, and digital factories—need development to establish tailored AI platforms and intelligent systems.

Encouraging leading industry enterprises to share data interfaces and create industry-level API standards, such as component data and standards, will promote interenterprise communication norms.

Moreover, the role of industry associations and sector-specific media cannot be underestimatedFor example, the promotion association is currently planning to collaborate with various ecological partners to establish 20 "AI Research-Driven Cost Reduction Collaborative Centers," aimed at uniting resources into concerted efforts while facilitating the substantial deployment of AI technologies within Guangdong’s manufacturing landscape.

These collaborative centers will serve as bridges for the deep integration of AI technologies with manufacturing, offering comprehensive services such as technical support, personnel development, and project incubations to assist in upgrading Guangdong's manufacturing sectorThe goal is to achieve large-scale applications of AI technology in manufacturing, creating a number of exemplary AI-driven manufacturing projects with demonstrable impacts.

Fourth, the promotion of low-code AI platforms, like Huawei's ModelArtsLite, can prove indispensableBy allowing engineers to customize models through a drag-and-drop interface and deploy edge knowledge nodes for localized real-time inference, it aids in constructing an industrial knowledge base that standardizes parameters within niche domains while encapsulating industrial wisdom, lowering algorithm development thresholds.

Selecting high-value scenarios while constructing refined knowledge bases will be critical

Industry-level knowledge-sharing platforms must be established to convert dispersed crafting expertise, equipment parameters, and fault cases into iterable digital assets, facilitating AI's roles as a "knowledge excavator," "structural engineer," and "intelligent applicator."

Creating an industrial knowledge base transcends mere technical engineering; it's fundamentally about reshaping productive relationshipsBy leveraging AI to convert the “sparks of knowledge from seasoned professionals” into “replicable digital ignite,” SMEs can transcend the constraints of experience transmission in space and time, allowing for a significant leap from “demographic dividends” to “intelligent dividends.”

Should Guangdong succeed in accomplishing this vital transformation, it could solidify its status as a “global knowledge hub for intelligent manufacturing.” The future produced here will not only involve tangible goods but also define the rules and standards for the manufacturing landscape ahead.

The AI transformation of SMEs is far from a straightforward technical issue; it fundamentally constitutes a comprehensive systemic engineering ventureIf Guangdong leverages the combinative strategies of “scenario-based technological provision + tiered cost-sharing + ecosystemic data governance” to surmount its obstacles, it may activate millions of SMEs as the capillaries for AI's deployment.

When every workshop can summon AI capabilities at minimal costs, regional competitiveness in manufacturing can evolve from a focus on “individual strength” to “systemic strength,” ultimately realizing a transformation similar to a swarm of ants overcoming challenges togetherThis transformation faces no retreat: industry foundations must be reconstructed using AI, or risks facing reconstructions themselves

Advertisements

Advertisements