Research on the Impact of Artificial Intelligence on the Resilience of the Manufacturing Industry Chain

Abstract

Artificial intelligence (AI) is of enormous significance for enhancing the resilience of the manufacturing industry chain, providing opportunities and momentum. We examine the impact of AI on the resilience of the manufacturing industry chain using a sample of listed manufacturing companies from 2011 to 2023. The results indicate that AI significantly improves the resilience of the manufacturing industry chain. Heterogeneity analysis reveals that the promoting effect of AI on manufacturing industry chain resilience is more pronounced in growth-stage enterprises, large-scale enterprises, enterprises in eastern regions, regions with high marketization levels, and financially distressed enterprises. Furthermore, mechanism tests indicate that AI enhances the resilience of the manufacturing industry chain by promoting firms’ ESG performance, facilitating knowledge spillovers, and increasing stock price synchronicity. The findings provide empirical evidence for the mechanisms and pathways to enhance the resilience of the manufacturing industry chain, offering insights into how AI can empower the high-quality development of China’s economy.

1. Introduction

The key to achieving economic growth lies in industrial development. Over two-thirds of global trade is conducted through the division of labor in industry and value chains, which has become the critical organizational model of modern manufacturing industry [1]. Today, the world is undergoing profound changes unseen in a century, with rising anti-globalization trends, intensified geopolitical uncertainties, and the combined impact of the COVID-19 pandemic and extreme policies disrupting the traditional international division of labor system. The security of the global industry chain faces severe challenges, and the contradiction between efficiency and stability on which their cooperation relies has become increasingly prominent. The industrial chain system urgently needs to shift from focusing solely on efficiency to “balancing efficiency and security.” The manufacturing sector constitutes the backbone of the national economy, serving as the foundation for state-building, the engine for national development, and the basis for national strength. The 20th National Congress of the Communist Party of China particularly emphasizes strengthening the resilience and security of industrial and supply chains. China’s industrial chain has long been positioned in the mid- to low-end segments of the global value chain, with relatively weak foundational industrial capabilities, reliance on foreign core technologies, and prominent “bottleneck” constraints. In certain areas, independent innovation capacity remains insufficient [2], making China’s manufacturing sector particularly vulnerable to external shocks in international competition. Enhancing the resilience of the manufacturing industry chains is therefore not only crucial for stabilizing the overall economy but also a key support for accelerating the realization of Chinese-style modernization.

Amid the wave of the new technological revolution and industrial transformation, AI has injected powerful momentum into enhancing the resilience of the manufacturing industry chain, profoundly transforming traditional production and operation models. Industrial intelligence plays a crucial role in reducing production costs and promoting the development of advanced technologies and has become a key pathway for enhancing industry and supply chain resilience and advancing China’s manufacturing sector toward the mid- to high-end of the global value chain [3]. In 2023, the scale of China’s core AI industry increased by 11.9% year over year, reaching 175.1 billion RMB. The 2024 AI Index Report released by Stanford HAI also highlighted significant global progress in AI in 2023, with 61 top AI models originating from the United States and 15 from China. Furthermore, the 2024 Government Work Report emphasized the need to deepen research and application of AI, big data, and other technologies, promoting the “AI+” initiative. Against the backdrop of the rapid growth of the intelligent industry and strong national emphasis on AI technologies, this raises several key questions: Will AI enhance the resilience of manufacturing industry chains? If AI has a significant impact, what are the underlying mechanisms? How do heterogeneous factors affect the extent to which AI strengthens industry chain resilience? We address these questions, building on existing research, to deepen the understanding of the socioeconomic consequences of AI and identify pathways to improve the resilience of manufacturing industry chain. Investigating the mechanisms and effects of AI on industry chain resilience can help the manufacturing sector better withstand risks, accelerate the healthy development of industry chain, and reinforce the foundational stability of the real economy.

2. Literature Review

The first stream of literature relevant to this study focuses on the effects of AI. Existing research indicates that AI exerts positive impacts on firm development through multiple channels. AI technologies can promote corporate innovation [4], enhance product innovation, support company growth, and facilitate the development of “star” firms [5], as well as mitigate the decline in innovation output following IPOs [6]. Moreover, the application of AI has enabled manufacturing firms to leverage human intelligence for green value creation [7]. Additionally, both AI exposure and board network heterogeneity independently reduce information risk; however, in industries where AI is most pervasive and critical, heterogeneous board networks may increase information risk and weaken the risk-reducing effect of AI exposure [8]. AI also drives total factor productivity (TFP) improvements by promoting R&D investment and optimizing the allocation of internal human capital within service firms [9], thereby enhancing overall productivity [10]. Furthermore, AI-enabled marketing positively affects firm profitability, customer satisfaction, and customer acquisition [11] and can improve firm performance through organizational and customer agility [12]. Chen et al. [13] observe that AI enhances firms’ ESG performance by increasing total factor productivity and R&D expenditures.

The second part of the literature pertains to research on the resilience of the manufacturing industry chain, primarily focusing on its conceptualization and influencing factors.

In terms of conceptualization, the term “resilience” originates from the Latin word resilio, which originally meant “to return to the initial state.” Initially, it was used to describe the ability of a system or individual to recover to its original state after external shocks or disturbances. However, since the 1970s, the connotation of resilience has been broadly expanded and developed. Today, resilience is more commonly understood as the capacity of a system to recover from adverse conditions, either returning to its original state or adapting to a new state that meets evolving demands or circumstances [14]. Lin et al. [15] define resilience as the adaptive capability of an industry chain to respond swiftly and restore its original connectivity, operational continuity, and control capacity when subjected to shocks, thereby effectively preventing and mitigating disruptions.

In terms of influencing factors, Cui et al. [16] demonstrate that new-type productivity enhances industry chain resilience by promoting urban–rural integration; however, government intervention weakens the effect of new-type productivity on industry chain resilience. Yan [17], using the open public data platform as a quasi-natural experiment, pointed out that open public data primarily strengthens industry chain resilience by improving government governance efficiency and stimulating innovation and entrepreneurship. Zheng et al. [18], based on data from 31 provinces in China from 2013 to 2022, concluded that a favorable business environment can enhance industry chain resilience by improving industry chain finance levels while also generating positive spillover effects on neighboring provinces. Li et al. [19] observe that the establishment of intellectual property demonstration cities can significantly enhance urban industrial diversification and innovation capacity, thereby improving the resilience of urban industry chains. Zhou et al. [20], using “supplier–customer” data from resource-based enterprises, identify a significant U-shaped relationship between government subsidies and the resilience of the resource-based industry chain, ultimately enhancing overall industry chain resilience. Yan et al. [21], using the new energy vehicle industry as an example, observe that AI penetration exerts an inverted U-shaped effect on the shock resistance, recovery, and re-innovation capabilities of the industry chain, as well as on overall resilience, although the average effect has not yet surpassed the turning point. Furthermore, studies have demonstrated the positive effects of digital infrastructure construction [22] and smart logistics [23] on industry chain resilience.

Existing literature has made significant contributions to understanding the relationship between AI and the resilience of manufacturing industry chains. However, several limitations remain. Some studies integrate AI and manufacturing industry chain resilience into a unified framework, but they focus primarily on the macro level or specific manufacturing sectors, lacking a comprehensive micro-level analysis. This study potentially contributes to three main aspects. First, from the research perspective, existing studies largely rely on macro-level analyses at the regional or industry level. This study, using A-share listed manufacturing firms as the sample, establishes a micro-level causal chain linking AI to industry chain resilience. Through rigorous econometric testing, it confirms the direct positive effect of AI on industry chain resilience, providing a solid microfoundation for understanding technology-enabled resilience. Second, from the theoretical mechanism perspective, this study overcomes the limitations of single-path explanations by constructing a multidimensional transmission framework. We, based on Resource-Based Theory, Dynamic Capabilities Theory, Stakeholder Theory, Corporate Reputation Theory, Organizational Learning Theory, and the Theory of Information Spillover Effects, reveals the mechanism by which AI enhances the resilience of the industry chain through three parallel channels: improving ESG performance, promoting knowledge spillover, and enhancing stock price synchronicity. This multidimensional framework surpasses the traditional single-perspective approach and comprehensively demonstrates the complex process through which technological capabilities translate into organizational resilience. Thirdly, from a practical guidance perspective, this study transcends universal conclusions and elucidates the context-dependent nature of technology-enabled effects. Systematic heterogeneity analysis shows that AI’s enabling effect is more pronounced in growth-stage firms, large-scale firms, firms located in eastern regions, regions with high marketization, and financially distressed firms. These findings not only clarify the boundary conditions of technological applications but also provide a theoretical basis for firms to develop differentiated transformation strategies and for governments to implement targeted industrial policies, avoiding blind promotion of technology adoption.

3. Theoretical Analysis and Research Hypotheses

3.1. Direct Effect of AI on Enhancing Manufacturing Industry Chain Resilience

Resource-based theory suggests that the unique and hard-to-replicate internal resources and capabilities possessed by a company are the cornerstone of its sustainable competitive advantage and superior performance. As a valuable, inimitable, and irreplaceable core resource, AI is increasingly becoming an important component of a company’s competitive strength. AI has permeated extensively into the production operations of businesses. Manufacturing enterprises, with the help of intelligent algorithms, can optimize production processes, restructure workflow, and enhance the flexibility of their production systems. For example, AI-driven Robotic Process Automation (RPA) technology can enable cross-system interactions, making complex tasks more controllable and reducing human errors [24], thereby improving the input-output efficiency of manufacturing enterprises.

Dynamic capabilities theory emphasizes that in the face of changes in the external environment, companies must possess the ability to rapidly adapt and adjust their resources. AI can predict potential risks such as supply chain disruptions or equipment failures and formulate proactive response strategies. In the event of unexpected disruptions, intelligent scheduling systems can quickly replan production schedules, allocate materials, and adjust logistics paths to minimize interference, ensuring the continuity and stability of critical production processes. In addition, by analyzing large volumes of user data, manufacturing firms can significantly improve the efficiency of matching supply and demand, establish flexible and responsive production models tailored to actual user needs, and further explore and expand product functionalities. This not only enhances firm competitiveness but also drives the horizontal expansion of the manufacturing industry chain, creating new competitive advantages within the industrial ecosystem [25].

The synergistic effects of AI not only enhance a company’s ability to integrate technology and resources but also promote cross-industry and cross-regional collaboration. On one hand, AI drives the efficient flow of data and information within manufacturing enterprises, breaking down information barriers between production stages, overcoming spatial and temporal limitations, and facilitating the efficient integration of dispersed resources. It accelerates the flow of production factors and provides more diversified paths for the allocation of labor, technology, data, and other resources. On the other hand, AI alleviates the limitations of geographical distance on cross-regional collaboration among manufacturing enterprises, fostering closer connections and more frequent interactions between the various stakeholders in the industry and supply chains. Enterprise innovation models are gradually shifting from traditional, closed, independent approaches to open, collaborative innovation. In particular, AI encourages affiliated firms to jointly tackle critical core technologies, effectively addressing “bottleneck” problems in industry and supply chains [26], and promoting the deep integration of data chain, innovation chain, talent chain, and capital chain, and ultimately strengthens the resilience of manufacturing industry chains. Based on the above analysis, we propose Hypothesis 1:

H1. 

AI can enhance the resilience of the manufacturing industry chain.

3.2. The Indirect Effects of AI on Enhancing the Resilience of Manufacturing Industry Chains

3.2.1. Corporate ESG

Stakeholder theory suggests that a company’s sustainable development requires balancing and meeting the expectations and interests of all stakeholders, not just shareholders. AI, as a tool supporting sustainable development, plays a significant role in improving corporate ESG performance. From the environmental dimension (E), AI leverages its unique platform effect and supervisory constraint effect to play a crucial role in ecological and environmental protection. Its platform effect aggregates massive environmental data and advanced algorithm models, providing solid technical support for precise control of pollutants and carbon emissions. The supervisory constraint effect enables real-time monitoring and early warning of pollution emissions during production processes, ensuring effective implementation of environmental protection measures. This significantly improves the accuracy of pollution emission control, reduces emission intensity, and injects strong momentum into promoting green and low-carbon development [27,28]. Meanwhile, through technological monitoring and intelligent tracking, environmental protection departments can accurately obtain corporate carbon emission data, effectively alleviating information asymmetry between environmental monitoring departments and enterprises. Based on authentic and reliable carbon emission information, governments implement “rewarding excellence and punishing inferiority” policies to incentivize enterprises to actively reduce emissions and promote green development [27], thereby further optimizing the environmental performance of enterprises.

From the social dimension (S), enterprises utilize real-time monitoring and intelligent detection systems during production to promptly identify potential issues in production processes and make automatic adjustments. This avoids human operational errors, reduces product defects, ensures product consistency and reliability, enhances consumer trust and satisfaction, and helps elevate corporate image and social responsibility. Furthermore, through processing and analyzing large amounts of data, AI enables enterprises to accurately predict market trends and optimize product designs and service models, thereby better meeting the expectations of different stakeholders. Such technology not only improves communication efficiency between enterprises and stakeholders but also fosters positive interactions between enterprises and society. Simultaneously, enterprises use AI to enhance employee training and development, improving employees’ skill levels and professional competencies, thereby promoting social employment and economic development.

From the governance dimension (G), AI technologies have enabled the automation and intelligent processing of corporate information disclosure. AI can rapidly extract and analyze both financial and non-financial information, effectively reducing human manipulation and information delays, thereby significantly enhancing the accuracy, completeness, and transparency of disclosures. At the same time, AI-driven intelligent decision-making systems provide management with scientific decision support, enabling the identification of risks, evaluation of alternatives, and formulation of optimal governance strategies in complex environments, thus improving the efficiency and quality of corporate governance. Moreover, AI plays a critical role in internal auditing and compliance management. Firms can leverage AI to detect anomalous transactions, identify financial fraud, and monitor anti-corruption risks, achieving more automated and precise internal controls, which enhances risk prevention and organizational integrity.

Strong ESG performance can enhance corporate reputation, alleviate financing constraints, and improve employee cohesion, thereby boosting the resilience of manufacturing industry chains. First, reputation is an intangible asset accumulated through long-term corporate operations. According to corporate reputation theory, a satisfactory reputation enables enterprises to gain widespread trust and recognition across society, thereby providing strong support for acquiring scarce and hard-to-replicate valuable resources [29]. Strong ESG performance can enhance corporate credibility, increase market recognition, win trust from external stakeholders, attract high-quality partners, establish stable cooperative relationships, and secure resources needed for corporate development. These resources and opportunities can serve as “insurance” during crises, helping enterprises cope with external shocks and enhancing the resilience of manufacturing industry chains. Furthermore, strong ESG performance reflects corporate emphasis on employee welfare and active promotion of diversity and inclusive culture, which can enhance organizational adaptability and stimulate employee motivation and creativity [30]. Simultaneously, it can attract external top talent, create a virtuous cycle of talent aggregation, and further strengthen team cohesion [31]. As a positive organizational identity symbol, ESG can enhance employees’ sense of belonging to the enterprise, motivating them to work together for long-term corporate development, thereby optimizing individual organizational behavior and work performance. During crises, this sense of belonging and cohesion can significantly improve corporate recovery capability and resilience [32]. Based on this, we propose Hypothesis 2:

H2. 

AI enhances the resilience of manufacturing industry chains by improving the ESG performance of manufacturing enterprises.

3.2.2. Knowledge Spillover

Organizational Learning Theory posits that an organization is not simply a collection of individuals, but a system capable of acquiring, creating, and transmitting knowledge, and adjusting its behavior accordingly, much like an organism. Its ultimate goal is to enhance the organization’s competitive ability to adapt to environmental changes. By establishing systematic knowledge transmission mechanisms, AI technologies have profoundly transformed the way manufacturing firms share knowledge. Organizational boundaries, information barriers, and limited transmission channels traditionally constrained knowledge flows, leading to slow dissemination, significant information loss, and low collaboration efficiency. The application of AI effectively overcomes these limitations by constructing a multi-level knowledge governance system, enabling efficient knowledge circulation and rational allocation. Specifically, in terms of knowledge conversion, AI-driven deep learning systems can extract and organize tacit knowledge embedded in manufacturing processes, transforming expertise in process optimization, equipment diagnostics, and other individually dependent knowledge into standardized, reusable information. In terms of knowledge matching, intelligent algorithms, utilizing enterprise capability maps and demand profiles, can identify knowledge gaps and potential suppliers within the industry chain, establishing precise supply-demand connections. Finally, regarding knowledge collaboration, cloud-based digital work environments facilitate real-time interactions among firms, allowing cross-temporal and cross-spatial joint innovation and value co-creation, gradually forming an open, efficient, and sustainable knowledge-sharing network that promotes inter-firm knowledge spillovers.

Knowledge spillovers generate positive externalities [33], and improvements in inter-firm knowledge-sharing efficiency provide a foundation for enhancing industry chain resistance and recovery. In terms of resistance, the widespread circulation of knowledge encourages the decentralized distribution of critical technologies and production capabilities, enabling the industry chain to rely on multi-node capacity reserves in response to external shocks, thus avoiding cascading failures from single points of disruption and ensuring the continuous operation of core processes. For example, key technologies being mastered by multiple firms provide technological redundancy, while the dispersed distribution of core production capabilities establishes backup capacity, enhancing overall risk resistance. In terms of recovery, an efficient knowledge network supports firms in rapidly responding to unexpected events. Leveraging intelligent knowledge retrieval and recommendation systems, firms can quickly access solutions for fault handling, process adjustment, and industry chain reorganization, significantly shortening emergency response times. More importantly, this mechanism creates a continuous learning loop: each crisis experience is codified and transformed into improved contingency plans and operational protocols, driving the industry chain to continuously refine its crisis response patterns. As a result, the industry chain demonstrates greater adaptability and self-repair capacity in the face of various challenges, achieving sustained resilience and long-term development. Based on the above, we propose Hypothesis 3:

H3. 

AI enhances the resilience of manufacturing chains by promoting knowledge spillovers.

3.2.3. Stock Price Synchronicity

Stock price synchronicity essentially refers to the phenomenon where individual stock price movements tend to converge, meaning that stock prices reflect more of the overall market trends rather than firm-specific information. In a context where market mechanisms are more effective and corporate governance is more advanced, individual stock prices are more likely to fluctuate in line with the overall market, resulting in higher price synchronicity [34].

Information spillover effects theory refers to the idea that the impact of an organization’s actions can extend beyond its direct goals and influence external entities that are not directly involved, creating unintended ripple effects. These external impacts can be both constructive and destructive, meaning they can result in both positive and negative externalities. In traditional capital markets, problems such as low-quality, delayed, and fragmented corporate information disclosure make it difficult for investors to fully and promptly grasp the real operating conditions of firms along the industry chain. As a result, market prices fail to reflect systematic information fully. AI, however, can automatically identify, extract, and integrate multi-source information—including corporate financial reports, announcements, media coverage, and social media content—thereby greatly enhancing the transparency and consistency of information disclosure. This improvement in information processing enables the capital market to form more convergent assessments of risks and values among firms in the industry chain, strengthening the co-movement of their stock prices and resulting in a higher level of stock price synchronicity. Meanwhile, AI also promotes a shift in investment decision-making from individual-based to system-based approaches through algorithmic trading and intelligent investment advisory services. This transformation leads to a more unified perception of the overall performance of the industry chain among market participants. Consequently, valuation biases caused by information asymmetry among firms are reduced, and stock prices better reflect macro trends and systemic risks at the industry chain level. Moreover, with its powerful information integration capabilities, AI can to some extent predict adverse events for firms. When such predicted events occur, they tend to cause fewer unexpected shocks to stock prices. As a result, today’s stock prices—containing more information—are expected to include less firm-specific information about the future and exhibit higher synchronicity [35]. Therefore, stock price synchronicity serves as an external manifestation of AI’s role in capital markets, reflecting improvements in the efficiency of information flow and the coordination of market expectations within the industry chain.

Higher stock price synchronicity indicates stronger information linkages among firms along the industry chain and enables the market to identify potential systemic risks more promptly. When fluctuations occur at one stage of the chain, price signals in the capital market quickly transmit to upstream and downstream firms, prompting timely adjustments in production, inventory, and investment decisions—thus preventing local risks from spreading into full-chain shocks. Furthermore, from the perspective of resilience, the efficient information-sharing mechanism reflected by stock price synchronicity allows the capital market to reallocate resources such as capital, technology, and labor more swiftly to the segments most affected by shocks, helping firms restore production capacity and shorten the recovery cycle of the industry chain. At the same time, the embedded application of AI in the capital market makes risk warning, investment evaluation, and policy response more intelligent and forward-looking, forming a dynamic adjustment and coordinated defense mechanism at the industry chain level. Based on the above, we propose Hypothesis 4:

H4. 

AI enhances the resilience of the manufacturing industry chain by improving stock price synchronicity.

4. Research Design

4.1. Econometric Model Specification

To test Hypothesis 1, we construct the following baseline regression model:

$${\mathrm{R}\mathrm{e}\mathrm{s}\mathrm{i}\mathrm{l}}_{\mathrm{i},\mathrm{t}}={\mathsf{\alpha }}_0+{\mathsf{\alpha }}_1{\mathrm{A}\mathrm{I}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\alpha }}_2{\mathrm{X}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{\delta }}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\mathrm{i},\mathrm{t}}$$

(1)

In the model, AI_i,t serves as the independent variable, representing AI in manufacturing; Resil_i,t is the dependent variable, indicating the resilience of the manufacturing industry chain; X_i,t represents a set of control variables that may influence the resilience of the manufacturing industry chain; μ_i denotes individual fixed effects to account for time-invariant firm-specific characteristics; δ_t represents time fixed effects to control for common annual shocks; and ε_i,t is the random disturbance term. In Model (1), the primary focus is on the sign and significance of the regression coefficient α₁. If α₁ is significantly positive, it validates Hypothesis 1, indicating that AI effectively enhances the resilience of the manufacturing industry chain. Conversely, if α₁ is significantly negative, it suggests that AI has a dampening effect on industry chain resilience.

4.2. Variable Definition

4.2.1. Dependent Variable

Resilience of the Manufacturing Industry Chain (Resil). Referring to the study by Li et al. [36], this paper constructs an evaluation index system for the resilience of the manufacturing industry chain in terms of resistance and recovery. The resilience of the manufacturing industry chain is then measured using the entropy weight method, and the resulting scores are multiplied by 100.

Table 1. Measurement Index System of Manufacturing Industry Chain Resilience.

Primary Dimension	Secondary Indicator	Definition and Calculation
Resistance	Bullwhip Effect	Degree of supply–demand fluctuation deviation
	Customer Concentration	Sales amount of the top five customers/Total sales
	Cash Holding Ratio	(Monetary funds + Trading financial assets)/Total assets
	Current Assets to Liabilities Ratio	Current liabilities/Total assets
	Inventory Turnover Ratio	Number of inventory turnover days
Recovery	Supplier Concentration	Procurement amount from top five suppliers/Total procurement
	Long-term Solvency	Equity ratio
	Long-term Assets to Liabilities Ratio	Long-term liabilities/Total assets
	R&D Investment	R&D expenditure/Operating revenue

displays the specific indicators.

4.2.2. Independent Variable

AI. Referring to the approach of Liu et al. [37], the level of AI of manufacturing firms is measured by the natural logarithm of one plus the number of AI-related patents applied for by the listed company in the current year.

4.2.3. Mediating Variables

Corporate ESG Performance: This study uses the ESG rating system developed by Huazheng ESG Rating, which assigns companies a rating from AAA to C, comprising a total of nine levels, updated quarterly. For quantitative analysis, these levels are converted into scores ranging from 9 (highest) to 1 (lowest), and the average of the four quarterly scores is used to represent a firm’s ESG performance in a given year. Knowledge Spillover: Knowledge spillover mainly occurs through the mobility of skilled personnel. Following the study of Zhang et al. [38], this paper measures the knowledge spillover effect of manufacturing firms by the natural logarithm of the number of R&D personnel employed by the enterprise. Stock Price Synchronicity: Referring to the method of Durnev et al. [39], this paper estimates the R² of individual stocks using model (2) and then applies Equation (3) to log-transform R² to obtain a normally distributed variable. The resulting indicator, Syn, serves as the measure of stock price synchronicity.

$${\mathrm{r}}_{\mathrm{i},\mathrm{t}}={\mathsf{\beta }}_0+{\mathsf{\beta }}_1{\mathrm{r}}_{\mathrm{m},\mathrm{t}}+{\mathsf{\beta }}_1{\mathrm{r}}_{\mathrm{I},\mathrm{t}}+{\mathsf{\epsilon }}_{\mathrm{i},\mathrm{t}}$$

(2)

$${\mathrm{S}\mathrm{y}\mathrm{n}}_{\mathrm{i}}=\mathrm{L}\mathrm{n}\left({\displaystyle \frac{{\mathrm{R}}_{\mathrm{i}}^2}{1-{\mathrm{R}}_{\mathrm{i}}^2}}\right)$$

(3)

where ${\mathrm{r}}_{\mathrm{i},\mathrm{t}}$ denotes the weekly return of stock i in week t, ${\mathrm{r}}_{\mathrm{m},\mathrm{t}}$ represents the weekly market return, and ${\mathrm{r}}_{\mathrm{I},\mathrm{t}}$ is the weekly industry return. The variable ${\mathrm{r}}_{\mathrm{I},\mathrm{t}}$ is calculated based on the China Securities Regulatory Commission (CSRC) industry classification standard, using the circulating market value of each company as the weight to compute the weighted average of ${\mathrm{r}}_{\mathrm{i},\mathrm{t}}$. ${\mathrm{R}}_{\mathrm{i}}^2$ represents the goodness of fit of model (2).

4.2.4. Firm-Level Controls

This study includes several control variables: Proportion of Independent Directors (Indep): Calculated as the number of independent directors divided by the total number of board members. Audit Opinion (AO): Assigned a value of 1 if the audit opinion is unqualified, and 0 otherwise. Firm Size (Size): Measured by the natural logarithm of total assets. Ownership Balance (Rest_r): Calculated as the ratio of the combined shareholding of the second to fifth largest shareholders to that of the largest shareholder. Board Size (Board): Measured by the natural logarithm of the number of board members. Property Ownership (Soe): If the firm is a state-owned enterprise, assign a value of 1; otherwise, assign 0. CEO Duality (Dual): Assigned a value of 1 if the chairman and CEO are the same person, and 0 otherwise. The variable descriptions are shown in

Table 2. Variable Descriptions.

Variable Category	Variable Name	Variable Symbol	Variable Definition
Dependent Variable	Manufacturing Industry Chain Resilience	Resil	Entropy weight method
Independent Variable	Artificial Intelligence	AI	Natural logarithm of the number of AI-related patents plus 1
Firm-Level Controls	Firm Size	Size	Natural logarithm of total assets of the manufacturing firm
	Proportion of Independent Directors	Indep	Number of independent directors/total number of directors
	Audit Opinion	AO	Assigned 1 if unqualified opinion, otherwise 0
	Ownership Balance	Rest_r	Sum of shareholding ratios of the 2nd to 5th largest shareholders/shareholding ratio of the largest shareholder
	Board Size	Board	Natural logarithm of the number of board members
	Property Ownership	Soe	Assigned 1 if state-owned enterprise, otherwise 0
	CEO Duality	Dual	Assigned 1 if the chairman and CEO are the same person, otherwise 0
Mediating Variables	Corporate ESG	ESG	Huazheng ESG Rating
	Knowledge Spillover	Kom	Natural logarithm of the number of R&D personnel
	Stock Price Synchronicity	Syn	Calculated using Formulas (2) and (3)

.

4.2.5. Sample Source and Descriptive Statistics

This study uses data from A-share listed manufacturing firms in China for the period 2011–2023 as the research sample. The sample is refined and processed as follows: ST-designated firms and observations with missing key variables are excluded; for a small number of missing values, the linear interpolation method is applied. The data are sourced from the CSMAR database and the Wind database. Descriptive statistics of all variables are presented in

Table 3. Descriptive Statistics of Variables.

Variable	Observations	Mean	SD	Minimum	Maximum
Resil	24,278	5.658	1.362	1.381	88.350
AI	24,278	0.386	0.794	0.000	6.806
Size	24,278	22.07	1.176	18.39	27.640
Board	24,276	2.100	0.193	1.386	2.890
Indep	24,278	0.377	0.055	0.333	0.667
Dual	24,278	0.345	0.475	0.000	1.000
Rest r	24,278	0.795	0.624	0.004	4.000
AO	24,278	0.979	0.142	0.000	1.000
Soe	24,278	0.250	0.433	0.000	1.000
ESG	24,278	4.174	0.894	1.000	7.750
Kom	24,278	5.433	1.361	0.000	11.540
Syn	24,278	−0.814	1.061	−10.980	2.269

.

5. Results and Analysis

5.1. Benchmark Regression

Table 4. Benchmark Regression Results.

Variable	(1)	(2)	(3)	(4)
	Resil	Resil	Resil	Resil
AI	0.181 ***	0.027 **	0.176 ***	0.029 ***
	(0.011)	(0.011)	(0.011)	(0.011)
Size			0.045 ***	−0.045 ***
			(0.008)	(0.017)
Board			−0.107 *	0.069
			(0.060)	(0.071)
Indep			0.538 ***	0.419 *
			(0.206)	(0.216)
Dual			0.071 ***	0.028
			(0.020)	(0.020)
Rest_r			0.062 ***	0.049 **
			(0.015)	(0.021)
AO			0.219 ***	0.080 *
			(0.064)	(0.046)
Soe			−0.186 ***	−0.141 ***
			(0.023)	(0.044)
Firm & Year Fixed Effects	NO	YES	NO	YES
Constant	5.588 ***	5.642 ***	4.386 ***	6.246 ***
	(0.010)	(0.007)	(0.242)	(0.412)
Observations	24,278	24,278	24,278	24,278

Notes: *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively. Robust standard errors are reported in parentheses. The same notation applies to the tables below.

reports the baseline regression results of AI for the resilience of the manufacturing industry chain (Resil). Column (1) presents the regression results without the inclusion of control variables, individual fixed effects, or time fixed effects. Based on this, Column (2) adds individual and time-fixed effects. Column (3) further includes a series of control variables. Finally, Column (4) incorporates both control variables and fixed effects. The results across Columns (1) to (4) consistently show that the coefficient of AI remains significantly positive at least at the 5% significance level, regardless of whether fixed effects or control variables are included. This provides empirical support for Hypothesis 1.

5.2. Robustness Tests

5.2.1. Replacement of the Independent Variable

In this study, the number of AI-related patents filed in a region in the current year plus one, logarithmically transformed (Lnpatents_region), and the level of firm AI investment (AI_investment) are used as alternative variables for manufacturing industry AI to conduct robustness tests. As shown in Columns (1) and (2) of

Table 5. Robustness Tests: Alternative Core Variables, High-Dimensional Fixed Effects, Additional Controls, and Sample Adjustment.

	(1)	(2)	(3)	(4)	(5)	(6)
Variable	Alternative Core Variables		City FE	City × Year FE	Additional Controls	Sample Adjusted
Lnpatents_region	0.060 *
	(0.031)
AI_investment		0.015 **
		(0.006)
AI			0.028 **	0.031 ***	0.043 ***	0.025 **
			(0.011)	(0.011)	(0.014)	(0.011)
Controls	YES	YES	YES	YES	YES	YES
Firm & Year Fixed Effects	YES	YES	YES	YES	YES	YES
Observations	24,278	24,278	24,278	24,278	24,278	19,641

Notes: *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively. Control variables and constant terms are omitted for brevity. The same applies to the tables below.

, the coefficients of Lnpatents_region and AI_investment are both significantly positive, indicating that the conclusion that firm AI promotes the resilience of the manufacturing industry chain remains robust after replacing the core explanatory variable.

5.2.2. High-Dimensional Fixed Effects

To alleviate potential biases caused by geographic differences and other unobservable time-varying factors, this study further controls for city fixed effects and city-by-year interaction effects to mitigate endogeneity concerns in the empirical analysis. The results are reported in Table 5. Columns (3) and (4) include city fixed effects and city × time interaction terms, respectively. The AI coefficients remain significantly positive with similar significance levels as in the baseline regressions, further validating the core conclusion of this paper.

5.2.3. Inclusion of Additional Control Variables

On top of the original control variables, this study adds other potential determinants of manufacturing industry chain resilience, including: Government intervention, measured by government fiscal expenditure as a percentage of GDP; Economic development level, proxied by city per capita GDP; Internal control, measured by the internal control index provided by DIBO Company; Cultural capital, measured by per capita public library holdings; Intellectual property protection, measured by the IP protection index from the National Intellectual Property Development Report; Financial development, measured by the ratio of annual year-end deposits and loans to regional GDP. Column (5) of Table 5 shows that after including these variables, the coefficient of AI remains significantly positive at the 1% level (0.043).

5.2.4. Sample Adjustment

Considering that municipalities directly under the central government (Beijing, Shanghai, Tianjin, and Chongqing) may exert unique influences on the relationship between AI and manufacturing industry chain resilience due to their specific policies and resources, these regions are excluded from the sample to reduce potential confounding effects. Column (6) in Table 5 reports the regression results after this sample adjustment, where the estimated coefficient of AI remains positive and statistically significant at the 5% level. This result suggests that even after excluding the municipalities, AI continues to effectively promote the resilience of the manufacturing industry chain.

5.3. Endogeneity Test

There may exist a certain degree of reverse causality between AI and the resilience of the manufacturing industry chain. To address potential endogeneity issues, this study employs two instrumental variables: the lagged value of the independent variable (IV1) and the regional mean level of AI among firms (IV2). For IV1, the lagged AI level is highly correlated with the current AI level because technological investment and application exhibit continuity. In terms of exogeneity, the lagged AI level, after controlling for other factors that may affect industry chain resilience, does not directly influence the current resilience of the manufacturing industry chain; its effect is transmitted only indirectly through the current AI level. For IV2, the regional mean AI level affects individual firms’ AI adoption through technological spillovers and competitive pressure, making it strongly correlated with the endogenous variable. Regarding exogeneity, after controlling for other variables, the regional mean AI level does not directly affect the resilience of individual firms’ industry chains; its effect operates solely through the firm’s own AI level. The two-stage least squares (2SLS) method is then applied to test the model, and the results are presented in

Table 6. Instrumental Variable Results.

	(1)	(2)	(3)	(4)
Variable	First Stage	Second Stage	First Stage	Second Stage
	IV1		IV2
AI		0.453 ***		0.308 ***
		(0.128)		(0.027)
IV1	0.120 ***
	(0.015)
IV2			0.942 ***
			(0.022)
Kleibergen–Paap rk LM Statistic	42,894		518.950
	[0.000]		[0.000]
Kleibergen–Paap rk Wald F Statistic	57.537		1765.364
	{16.380}		{16.380}
Controls	YES	YES	YES	YES
Firm & Year Fixed Effects	YES	YES	YES	YES
Observations	19,590	19,590	24,278	24,278

Note: *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively. Values in [ ] represent p-values; values in { } represent the critical values for the Stock-Yogo weak instrument test at the 10% significance level.

. In the first-stage regressions, columns (1) and (3) show that the coefficients of the instruments IV1 and IV2 are positive and significant at the 1% level, satisfying the relevance condition. Columns (2) and (4) present the second-stage results. The Kleibergen–Paap rk LM statistic and the Kleibergen–Paap rk Wald F statistic indicate that both instruments pass the weak instrument and under-identification tests. Even after accounting for endogeneity, the AI level still significantly enhances the resilience of the manufacturing industry chain, further confirming the robustness of the results in this study.

6. Further Analysis

6.1. Mechanism Testing

Based on theoretical analysis and research hypotheses, this study investigates the mechanisms through which AI affects the resilience of the manufacturing supply chain, focusing on firm ESG performance, knowledge spillovers (Kom), and stock price synchronicity (Syn). Building on Model (1), a mediation effect model is constructed to test whether the proposed mechanisms hold.

$${\mathrm{E}\mathrm{S}\mathrm{G}}_{\mathrm{i},\mathrm{t}}={\mathsf{\alpha }}_0+{\mathsf{\alpha }}_1{\mathrm{A}\mathrm{I}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\alpha }}_2{\mathrm{X}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{\delta }}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\mathrm{i},\mathrm{t}}$$

(4)

$${\mathrm{K}\mathrm{o}\mathrm{m}}_{\mathrm{i},\mathrm{t}}={\mathsf{\alpha }}_0+{\mathsf{\alpha }}_1{\mathrm{A}\mathrm{I}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\alpha }}_2{\mathrm{X}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{\delta }}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\mathrm{i},\mathrm{t}}$$

(5)

$${\mathrm{S}\mathrm{y}\mathrm{n}}_{\mathrm{i},\mathrm{t}}={\mathsf{\alpha }}_0+{\mathsf{\alpha }}_1{\mathrm{A}\mathrm{I}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\alpha }}_2{\mathrm{X}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{\delta }}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\mathrm{i},\mathrm{t}}$$

(6)

The results in Column (1) of

Table 7. Mediation Analysis Results.

	(1)	(2)	(3)
Variable	ESG	Kom	Syn
AI	0.016 *	0.037 ***	0.023 *
	(0.009)	(0.010)	(0.012)
Controls	YES	YES	YES
Firm & Year Fixed Effects	YES	YES	YES
Observations	24,278	24,278	24,278

Notes: *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.

show that the coefficient of AI is positive and significant at the 10% level, indicating that enhancing firm ESG performance is a mechanism through which AI strengthens the resilience of the manufacturing industry chain. AI technology enables firms to implement more precise environmental management and more transparent corporate governance, thereby improving their overall ESG performance. This improvement enhances corporate reputation and internal cohesion, allowing firms to recover more quickly from external shocks and thereby increasing the stability of the entire industry chain, confirming Hypothesis 2. Column (2) presents the effect of AI on knowledge spillovers, and the results indicate a significant positive relationship. This suggests that AI can transform tacit knowledge into standardized information that can be shared across the industry chain, facilitating precise cross-organizational matching and collaborative R&D. Such knowledge flows enable the industry chain to mobilize technological reserves across multiple nodes when facing shocks, rapidly develop response strategies, and continuously strengthen resilience through accumulated crisis-handling experience, supporting Hypothesis 3. In Column (3), the AI coefficient is significantly positive, indicating that AI integrates multi-source information to make individual stock prices better reflect shared risks and value trends within the industry chain, forming more coordinated market expectations. When risks emerge, this coordination allows risk signals to be transmitted quickly, prompting firms to adjust decisions synchronously and guiding the efficient allocation of financial resources, thereby enhancing the overall risk resistance of the industry chain, confirming Hypothesis 4.

6.2. Heterogeneity Analysis

6.2.1. Heterogeneity by Firm Life Cycle

Drawing on the study by Dickinson [40], this paper divides the firm life cycle into three stages: growth, maturity, and decline. The sample is split accordingly for separate regression analyses. The results reveal significant differences in the effect of AI on manufacturing industry chain resilience across different life cycle stages. Specifically, the AI coefficient in Columns (1) of

Table 8. Heterogeneity Analysis by Firm Life Cycle.

	(1)	(2)	(3)
Variable	Growth Stage	Maturity Stage	Decline Stage
AI	0.031 **	0.007	0.026
	(0.013)	(0.051)	(0.032)
Controls	YES	YES	YES
Firm & Year Fixed Effects	YES	YES	YES
Observations	16,193	3402	4683

Notes: *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.

is 0.031 and significant at the 5% level for firms in the growth stage, 0.007 for firms in the maturity stage, and 0.026 for firms in the decline stage. The likely explanation is that firms in the growth stage are in a rapid expansion phase, urgently requiring technological innovation and efficiency improvements to support growth. AI significantly promotes firm growth by optimizing production processes, enhancing efficiency, and reducing costs. Firms in the maturity stage have established stable market positions and operating models, with relatively steady growth; AI mainly serves to maintain existing efficiency rather than generating substantial additional growth. For firms in the decline stage, resource constraints due to tight finances and talent loss, coupled with outdated production systems and poor data quality, reduce technology adaptability and thus limit the effectiveness of AI.

6.2.2. Heterogeneity by Firm Size

The talent structure, market share, and business models of firms vary with their size, which in turn causes differences in the impact of AI on the resilience of the manufacturing industry chain. This study measures firm size using the natural logarithm of the year-end total assets of manufacturing firms. The sample is divided into large-scale and small-scale enterprises based on the median value. Columns (1) and (2) of

Table 9. Heterogeneity Analysis by Firm Size and Region.

	(1)	(2)	(3)	(4)
Variable	Large Firms	Small Firms	Eastern Region	Central & Western Region
AI	0.038 **	0.010	0.043 ***	0.037
	(0.015)	(0.016)	(0.014)	(0.029)
Controls	YES	YES	YES	YES
Firm & Year Fixed Effects	YES	YES	YES	YES
Observations	11,451	12,827	17,213	7065

Notes: *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.

indicate that the positive effect of AI on enhancing industry chain resilience is more pronounced in large-scale manufacturing firms. The underlying reason may lie in the high financial costs associated with constructing and maintaining AI systems. Moreover, realizing the potential benefits of AI relies heavily on substantial intangible investments, such as data governance and talent training. In contrast, small- and medium-sized enterprises often face financial constraints, making it difficult to bear these upfront costs. In particular, SMEs encounter considerable uncertainty when evaluating the development costs of AI systems and their potential returns. Furthermore, the implementation of AI technologies does not typically yield immediate productivity gains; instead, it may initially generate sunk costs, which is particularly disadvantageous for resource-constrained SMEs. More importantly, SMEs are often inadequately prepared to extract value from data. Although they generate and process large volumes of data in daily operations, they generally lack the capacity to integrate, manage, and protect these data effectively. Compared with large enterprises, SMEs’ data may also be limited in both scale and quality, making it challenging to support high-quality AI model training and decision-making [41].

6.2.3. Heterogeneity by Region

China’s regions differ in terms of factor endowments and other conditions, and such regional differences may influence the extent to which AI enhances the resilience of the manufacturing industry chain. Based on this, the sample is divided into firms located in the eastern region and firms in the central and western regions. The results in Columns (3) and (4) of Table 9 indicate that the AI coefficient for manufacturing firms in the eastern region is 0.043 and significant at the 1% level, whereas the AI coefficient for firms in the central and western regions is 0.037 but not significant. This discrepancy may be attributed to the eastern region’s strong industrial foundation and superior factor endowments, including a more complete industrial supporting system, richer data resources, and higher-quality labor. The digitalization and intelligent transformation of manufacturing in this region started earlier, enabling firms to more effectively embed AI technologies across production and operational processes, thereby significantly enhancing industry chain resilience. By contrast, manufacturing in the central and western regions is largely resource-intensive or traditional processing, with weaker intelligent infrastructure, limited data access, and insufficient talent reserves, which constrain the absorption and application of AI technologies and make it difficult to achieve systemic improvements in the short term. Additionally, the eastern region benefits from a more optimized institutional environment and mature market mechanisms, providing firms with stronger incentives and capabilities to invest in AI, further amplifying its positive impact on industry chain resilience.

6.2.4. Heterogeneity by Marketization Levels

Differences in marketization levels may influence the extent to which AI can exert its effects. To examine this, the sample is divided into regions with higher and lower marketization levels based on the marketization index compiled by Fan Gang et al., and separate regressions are conducted for each group. The results in Columns (1) and (2) of

Table 10. Heterogeneity Analysis by Marketization Levels and Financial Distress.

	(1)	(2)	(3)	(4)
Variable	High Marketization Regions	Low Marketization Regions	Financially Distressed Group	Non-Financially Distressed Group
AI	0.023 *	0.025	0.037 ***	0.007
	(0.012)	(0.022)	(0.014)	(0.018)
Controls	YES	YES	YES	YES
Firm & Year Fixed Effects	YES	YES	YES	YES
Observations	13,645	10,633	15,405	8873

Notes: *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.

indicate that the impact of AI is more pronounced in regions with higher marketization levels, suggesting that a favorable market environment facilitates the realization of AI’s technological effects. Regions with higher marketization possess more efficient resource allocation mechanisms and a more open institutional environment. On the one hand, these regions feature stronger property rights protection and smoother factor mobility, providing sufficient innovation incentives and resources for AI development and application. Driven by competitive pressure and market signals, firms are more likely to leverage AI to enhance production efficiency and optimize decision-making processes, thereby improving both firm-level and supply chain resilience. On the other hand, regions with higher marketization have more developed financial systems and more transparent information disclosure, resulting in more positive market responses to AI-related investments, as well as stronger technology diffusion and learning effects. Furthermore, lower government intervention and a more inclusive institutional environment in these regions facilitate the deep integration of AI technologies into business operations, allowing their potential value in resource allocation optimization, risk identification, and management to be fully realized.

6.2.5. Heterogeneity by Financial Distress

For firms facing financial distress, AI plays a more significant role in improving operational efficiency, reducing costs, and exploring emerging business areas, thereby effectively helping firms to overcome their current financial difficulties. In highly competitive industries, the application of AI is likely a key strategy for firms to maintain a competitive advantage and improve market position. Accordingly, this study uses the risk score calculated by the O-Score model to measure the degree of financial distress, dividing listed firms into financially distressed and non-distressed groups based on the median score. Columns (3) and (4) of Table 7 indicate that AI has a stronger effect in the financially distressed group than in the non-distressed group. This may be because financially distressed firms face greater potential for efficiency improvement and urgent needs for change that traditional methods cannot meet. AI comprehensively optimizes production, sales, and other processes; reduces costs such as labor and raw materials; and explores new business opportunities. Moreover, financially distressed firms experience bottlenecks under traditional models, have stronger incentives for transformation, and place more emphasis on cost-effectiveness, making them more receptive to adopting AI and other new technologies. Conversely, non-distressed firms, due to path dependency from traditional advantages and balancing innovation investment with short-term benefits, may limit the full advantages of AI application.

7. Discussion

Due to limitations in the availability of micro-level data and the consistency of corporate disclosure standards, this study primarily focuses on A-share listed manufacturing firms in China to explore the impact of AI on industry chain resilience. However, this research design constrains the generalizability of the findings to non-listed enterprises, which constitute the majority of entities within industry chains. We infer that the mechanisms and effects of AI on industry chain resilience may vary substantially across firms of different sizes, resource endowments, and governance structures. For instance, large leading enterprises with abundant resources may leverage AI for deep data mining and intelligent decision-making, thereby significantly enhancing their industry chain control and risk resistance capabilities. In contrast, for numerous small and micro enterprises, the effects may be greatly diminished due to weak digital foundations, talent shortages, and financial constraints—or even exacerbate the “digital divide,” further weakening their bargaining power within the industry chain. Given these limitations, we can enhance and broaden future research in the following directions: Broaden and enrich the research sample. Subsequent studies may construct a mixed-sample database covering non-listed firms and small and micro enterprises. By collecting firsthand data through field surveys, questionnaires, and other empirical methods, researchers can explore the differentiated impacts of AI adoption and industry chain resilience building across firms of varying sizes and ownership types, thereby forming a more generalizable theoretical framework. Conduct cross-country and context-based comparative studies. Future research could select countries at different stages of development and with diverse institutional contexts to systematically analyze how market maturity, digital infrastructure, and industrial policies moderate the relationship between AI and supply chain resilience. This approach would help uncover the boundary conditions and contextual dependencies of AI’s impact on industry chain resilience. Focus on the inclusiveness and sustainability of technological applications. As AI technologies continue to advance, future studies should pay closer attention to the social implications of their deployment—such as the widening digital divide and shifts in employment structures—and explore how technology governance and policy guidance can ensure that AI development contributes to building a more inclusive and sustainable industry chain system.

8. Conclusions and Recommendations

8.1. Conclusions

This study conducts a comprehensive analysis of the effect of AI on the resilience of the manufacturing industry chain, utilizing a sample of A-share listed manufacturing firms in China from 2011 to 2023. The empirical findings show that AI significantly and consistently enhances resilience in the manufacturing industry chain, as confirmed by several robustness tests, highlighting AI’s crucial role in strengthening this resilience. Subsequent heterogeneity analysis indicates that the beneficial impact of AI on industry chain resilience is particularly significant among growth-stage enterprises, large firms, those in eastern regions, areas with elevated marketization, and those experiencing financial difficulties. Mechanism analysis reveals that AI bolsters the robustness of the industry chain by enhancing corporate ESG performance, promoting information spillovers, and augmenting stock price synchronization.

8.2. Recommendations

First, promote AI-enabled transformation in the manufacturing sector while balancing innovation-driven growth and cost constraints. AI is a key technology for enhancing the resilience of the manufacturing industry chain. It is essential to accelerate its application in production processes, industry chain management, and early-warning risk systems, thereby advancing the digital and intelligent transformation of the manufacturing industry. However, the high investment and maintenance costs associated with AI still pose significant barriers for small and medium-sized enterprises. Improper promotion may further exacerbate industrial polarization and technological dependence. At the policy level, a tiered implementation strategy should be adopted—reducing application costs through fiscal subsidies, shared technology platforms, and public data resources. Meanwhile, policymakers should avoid imposing uniform digital transformation requirements and instead ensure flexibility for firms with differing technological capacities. Internationally, cooperation should be strengthened in standard-setting and algorithm governance, taking into account differences in national development stages and adaptive capacities. Only by carefully assessing costs, capabilities, and potential risks can AI evolve into a sustainable driver of industry chain resilience, rather than a short-term technological burden.

Second, strengthen AI governance and its integration with corporate ESG practices to promote sustainable development in the manufacturing sector. It is essential to build a responsible, transparent, and sustainable AI governance system. On one hand, the government should improve ethical review mechanisms and data security regulations related to AI, and establish ESG-oriented evaluation standards for technological innovation. This ensures that AI applications not only enhance efficiency but also uphold fairness, environmental protection, and social welfare. On the other hand, enterprises should strengthen green production, energy conservation, emission reduction, and social responsibility during their digital transformation, embedding ESG principles into AI-based decision-making systems to achieve synergy between intelligent management and responsible governance. Furthermore, efforts should be made to align global AI ethics and ESG disclosure standards, fostering universally applicable green technology norms and cross-border cooperation frameworks. In this way, the empowerment of manufacturing by AI can simultaneously promote global sustainable development and industry chain stability.

Third, establish an international knowledge spillover mechanism to foster AI innovation and technology diffusion. Openness and knowledge sharing are essential pathways to enhancing the adaptability of the global manufacturing sector. China should leverage multilateral cooperation platforms such as the Belt and Road Initiative to build an international AI collaboration network, facilitating the cross-border flow of AI research and development, data resources, and innovation outcomes. The government can promote joint efforts among universities, research institutions, and enterprises to establish open laboratories and technology innovation centers, thereby forming mechanisms for cross-border R&D collaboration and talent cultivation. Meanwhile, it is crucial to improve intellectual property protection and result-sharing systems to encourage firms to open their algorithms and data interfaces under secure conditions, promoting technology diffusion and secondary innovation. Developed and emerging economies should also strengthen technological cooperation to narrow the gap in intelligent development and avoid the emergence of “technological islands.” Through such collaborative efforts, the global manufacturing system can enhance collective resilience and innovation capacity, ultimately forming a new pattern of global industry chain cooperation centered on AI.

Fourth, improve information systems for capital markets to increase financial resilience across the industry chain. The discovery that AI improves stock price synchronization implies that AI helps to improve the efficiency of information transmission and risk resistance in capital markets. It is consequently critical to fully harness AI in financial data analysis, risk monitoring, and investment decision-making, while also progressing the development of intelligent supervision and early-warning market systems. Governments and regulatory authorities can use AI technologies to improve the detection of aberrant market swings and cross-border capital flows, thereby reducing systemic financial risks. Meanwhile, businesses should increase information disclosure openness and use AI to better investor relations management and corporate value communication channels. At the international level, efforts should be made to encourage regulatory coordination in information disclosure, risk assessment, and data sharing across global capital markets, as well as the establishment of a global intelligent financial supervision framework aimed at reducing information asymmetry and market transmission delays. The development of an intelligent financial system has the potential to greatly improve both business financing resilience and the global industry chain’s financial stability and shock resistance.

Fifth, promote regional coordination and inclusive development to narrow the gap in intelligent application. To prevent regional imbalances in digital and intelligent transformation, national policies should strengthen regional coordination mechanisms and encourage the flow of AI technologies, capital, and talent toward central, western, and less-developed regions. Governments should increase investment in digital infrastructure for small and medium-sized enterprises and underdeveloped areas, establish regional AI application demonstration zones, and foster collaboration between local manufacturing firms, universities, and research institutes to enhance indigenous innovation and technology absorption capacity. Moreover, developing countries should be encouraged to participate in global intelligent manufacturing cooperation programs. Through technical assistance and policy support, they can better integrate into the global intelligent industry chain system and achieve inclusive growth. Only by narrowing the regional gap in intelligent development can a resilient, balanced, and globally competitive system for the manufacturing industry be truly established.