How Does Manufacturing Intelligentization Influence Innovation in China from a Nonlinear Perspective and Economic Servitization Background?

Abstract

Under the trend of high-quality economic development and economic servitization in China, it is of great significance to study the impact of manufacturing intelligentization on innovation and its mechanisms. This study uses a sample of 30 of China’s provinces from 2008 to 2020 to empirically test the effect of manufacturing intelligentization on innovation performance from a nonlinear perspective and examine the intermediary mechanism of diversified agglomeration of producer services against an economic servitization background. The study finds that intelligentization has a significant inverted U-shaped impact on innovation performance. It shows that the positive marginal intelligentization effect on innovation gradually decreases, and intelligence inhibits innovation when it exceeds the threshold. In addition, diversified agglomeration of producer services can improve innovation performance, and intelligentization has an inverted U-shaped effect on this diversified aggregation. Thus, the nonlinear influence of intelligence on innovation performance has a channel of diversified agglomeration of producer services. Furthermore, human capital has a reverse moderating effect on the inverted U-shaped relationship between intelligentization and innovation performance. This conclusion can help to promote the innovation-driven and sustainable development of China’s economy under the intelligent manufacturing strategy.

1. Introduction

With the development of manufacturing, China is rapidly growing into the world’s second-largest economy. However, China’s economy still faces severe difficulties in sustainable development. In the past, this country mainly relied on the low-cost advantage of production factors such as capital, labor, and energy. These advantages are constantly weakening, and the resource problems are gradually becoming more prominent. According to China’s National Bureau of Statistics, the proportion of China’s working-age population has slowly declined since 2010; from 2015 to 2020, the total working-age population showed a downward trend, decreasing by nearly 16.99 million people. Based on the BP Statistical Review of World Energy 2022, China’s primary energy consumption has been on the rise without restraint since it surpassed the United States in 2009. By 2021, China’s energy consumption was equivalent to 63.06 million tons of hard coal. This is about 1.7 times that of the United States and accounts for 26.5% of global primary energy consumption, ranking first worldwide. Moreover, China’s energy mining has been unable to meet its own development needs, and it has begun to import many energy and mineral resources from overseas. Thus, the lack of labor and the overconsumption of natural resources are gradually restricting China’s economic development. In light of the move toward the high-quality development of China’s economy, China’s extensive economic growth mode is no longer sustainable, and it is urgent to shift from factor-driven to innovation-driven mechanisms to promote sustainable economic development.

In the light of the Global Innovation Index from the World Intellectual Property Organization, China’s overall ranking of innovation capacity rose to 11th in the world in 2021, up 23 places from 2012. This is thanks to the in-depth implementation of China’s innovation-driven development strategy. However, China, with the world’s most enormous energy consumption and a large population, is far less efficient at innovation than developed regions such as Switzerland and the United States. China’s innovation performance still needs to be improved. At the same time, manufacturing intelligentization has become an important driving force for upgrading China’s manufacturing industry, innovation, and sustainable development [1]. “Made in China 2025” pointed out that intelligent manufacturing should be the main direction to promote the transformation and upgrading of the manufacturing industry. The 14th Five-Year Plan for Intelligent Manufacturing Development also indicated that the digital transformation, networking collaboration, and intelligent transformation of the manufacturing industry should be continuously promoted to provide strong support for the high-quality development of the manufacturing industry. Therefore, an in-depth study of the ability of intelligence to drive innovation is crucial to the high-quality and sustainable development of China’s economy.

Moreover, in implementing the Intelligent Manufacturing Strategy, in addition to the intelligent transformation of production, many large manufacturers are fundamentally changing their value creation strategies from product-centric to hybrid product–service providers [2,3]. Judging from international experience, the servitization of the economy is in line with the trend of a new round of technological revolution, industrial transformation, and consumption upgrading. Furthermore, it is also a practical choice for cultivating a new driving force for industrial development and a modern industrial system. In this context, enterprises are increasingly using services as intermediate inputs, resulting in the increasing externalization and marketization of producer services. Therefore, it is of great significance to the sustainable development of China’s economy to pay attention to the role of manufacturing intelligence and producer services in the innovation-driven strategy.

Recent research shows that implementing intelligent manufacturing is conducive to technological progress and regional innovation and is an essential measure for China’s innovation and development [4,5]. However, information technology in intelligent manufacturing has both an enabling effect and a squeezing effect on innovation, showing an inverted “U”-shaped curve relationship [6]. Therefore, the impact of manufacturing intelligentization on innovation is still unclear. Manufacturing intelligence is a vertically integrated production model that introduces the concept of the Internet of Things and servitization [7,8]. Under the trend of servitization transformation, can manufacturing intelligentization effectively promote technological innovation, thereby promoting China’s innovation-driven development? This problem remains to be empirically verified. If confirmed, what is the mechanism behind it? This study attempts to demonstrate the issues mentioned above in depth.

2. Literature Review

Research on the relationship between intelligence and innovation can be traced back to information technology and innovation literature. Information technology can improve the speed and efficiency of enterprise innovation through knowledge asset management, production support, and inter-organizational coordination [9]. Some studies also use empirical evidence to show that information technology can promote the output, performance, and process of innovation and new product development [9,10,11,12,13], or mitigate the diminishing returns to R & D [14]. However, if information technology is inflexible in the use process, it may lock enterprises in the situation of limited external knowledge sources, thereby inhibiting innovation [15]. Karhade and Dong (2021), based on the dynamic adjustment cost theory, found that with the increase in investment in information technology, the impact of information technology on innovation showed an inverted U-shaped relationship [6].

In the intelligent manufacturing system, intelligent technology will gradually replace human and mental activities, accelerating knowledge iterative updating and creation, improving learning and absorption, and promoting technological innovation [4]. Supported by big data, the development of deep learning technology will significantly reduce the cost of knowledge searching, prompting R & D departments to increase fixed capital investment in artificial intelligence. These investments may improve the performance of existing data-intensive research projects and open up new research ideas and development opportunities for studying social and physical phenomena previously outside systems science and empirical research [16]. Kakatkar et al. (2020) found through case studies that AI can leverage large-scale data for highly scalable and reproducible deep analysis, helping innovation teams validate creative insights, reduce creative blind spots, and uncover new problems in complex relationships [17]. Truong and Papagiannidis (2022) believe that artificial intelligence may acquire some creative ability by combining data in new ways to produce novel content, but whether it positively impacts disruptive innovation is unclear [18]. Grashof and Kopka (2022) further found that large companies increase radical innovation from artificial intelligence applications, while small and medium-sized enterprises use artificial intelligence technology as a general-purpose technology to promote fundamental innovation [19]. Rammer et al. (2022) also found the prominent role of AI in world-first innovations [20]. Grounded in gestalt insight learning theory and organizational learning theory, Ghasemaghaei and Calic (2019) studied the influence of big-data characteristics on enterprise innovation from the perspectives of data volume, data speed, data diversity, and data accuracy and found that the accuracy, speed, and diversity of big-data analysis are the keys to promoting enterprise innovation [21]. By increasing the scale and variety of information obtained by enterprises, reducing the cost of absorbing external knowledge, and promoting the integration of knowledge, big-data technology can improve effective results in the innovation process of enterprises [22]. Yang et al. (2022) believed that intelligence could produce a “technology promotion effect” and a “cost reduction effect”, promoting the level of regional green innovation [5].

Existing research has made some progress in the relationship between intelligent technology and innovation, which lays a theoretical foundation for subsequent research. However, there are still many problems to be further explored and solved. Firstly, the existing literature mainly studies the relationship between intelligent manufacturing and innovation performance with linear thinking, ignoring the potential negative impact on new knowledge acquisition caused by the excessive introduction of intelligent investment. It is necessary to explore the nonlinear relationship between manufacturing intelligence and innovation performance from a nonlinear perspective. Second, although the existing literature has recognized the trend of manufacturing servitization in the process of intelligent manufacturing, few scholars have included producer services in the study of the relationship between intelligence and innovation performance and have not investigated whether the change of producer service agglomeration exists in intelligence and innovation performance as a mediator. Therefore, the influence mechanism of manufacturing intelligence on innovation performance cannot be effectively revealed. Third, if there is a nonlinear influence of intelligence on innovation, few scholars discuss what measures can be taken to intervene in the extrusion effect of intelligence on innovation.

Based on this, the article uses the panel data of 30 provinces in China from 2008 to 2020 to conduct an empirical study on the impact of manufacturing intelligentization on innovation performance in a nonlinear way. In addition, from the perspective of diversified agglomeration of producer services, this study empirically adopts the mediation effect model to test the mechanism of manufacturing intelligentization on innovation performance. Thirdly, from the perspective of labor optimization, this paper uses the human capital variable to test its moderating influence on nonlinear effects.

Taking existing literature into consideration, the contribution of the study is mainly reflected by the following three points: (1) This study innovatively explores the nonlinear relationship between manufacturing intelligentization and innovation performance; (2) in terms of the analysis of mediating factors, this paper chooses the perspective of diversified agglomeration of producer services under the background of servitization; (3) this article also emphasizes the moderating effect of human capital on the nonlinear relationship to alleviate the marginal diminishing phenomenon of the impact of intelligentization on innovation performance. The research results can provide reliable suggestions for driving economic innovation and sustainable development in the era of intelligent manufacturing. The rest of the paper is structured as follows: Section 3 describes the research mechanism and the article’s hypotheses. Section 4 introduces the selection of variables and data for the empirical study, the research method, and the regression results. Section 5 discusses the results of the study. Section 6 concludes the study.

3. Theoretical Background and Research Hypotheses

3.1. Direct Mechanism

According to the research of innovation economics, technological innovation is the result of the reorganization of new and old knowledge and knowledge creation [23,24]. Correspondingly, the development of intelligent technology provides a new way to reveal new knowledge and view existing knowledge [4]. With the help of advanced intelligent technology, those manufacturers who have established cyber-physical systems (CPS) and carried out service transformation can obtain multivariate heterogeneous data of consumer terminals and the physical world. These massive amounts of data were not readily available before this advancement. Then, the manufacturers use image recognition techniques to synthesize the massive fuzzy image and text data and convert it into understandable and practical data, expanding the knowledge source. Intelligent technology can apply big-data analysis. The information management system of traditional enterprises can thus obtain the ability to process massive and cross-domain data [8]. Based on the acquired data, a large amount of explicit knowledge is brought to the enterprise through contextual correlation and predictive analysis. In addition, intelligent technologies can speed up identifying the highest value targets in the knowledge portfolio, providing more flexible and practical thinking for many vital technologies [21,25]. Thus, new knowledge generation and identification efficiency can be improved, and the product development cycle can be shortened. Intelligent systems achieve a certain degree of autonomous perception, autonomous learning, and autonomous decision-making capabilities through embedded big-data analysis, deep learning, and other technologies, which can continuously generate new knowledge [26,27]. Therefore, manufacturing intelligentization can facilitate technological innovation by improving data acquisition capabilities and accelerating knowledge identification and creation.

However, from the perspective of information processing theory, innovation may be inhibited at a highly intelligent manufacturing level. In this case, the information management system and information structure of the industrial chain or supply chain covering the “product–service” package will become large and complex. This trend can lead to information overload for individuals or organizations, increasing knowledge acquisition costs [28]. Specifically, the intelligent process of enterprises is the digital construction of the whole supply chain. The available digital information channels, information scale, and information diversity have increased rapidly (such as data from the Internet of Things and consumers). Due to the existence of bounded rationality, it is increasingly difficult for innovation workers to discover new knowledge from a large amount of information under limited processing power and attention span. This will reduce the efficiency of decision-making in R & D [6]. At a high level of intelligence, the region’s network of technical knowledge cooperation will be more closely linked. Enterprises can obtain a large amount of non-redundant knowledge from the network. However, inventors must also invest more resources and energy to identify, assimilate, transform, and utilize this knowledge [29,30]. Since the absorptive capacity of inventors is limited, when the knowledge contained in the accessible network exceeds a specific limit, it will also lead to an information overload of inventors, which is not conducive to innovation efficiency [31,32]. When inventors are overly dependent on a tight knowledge network, it may inhibit the inflow of external knowledge, create a technology lock-in effect among inventors, and limit the potential of inventors [33]. Based on the above analysis, this study proposes the research hypothesis H1:

H1. 

The impact of manufacturing intelligentization on innovation performance has an inverted U-shaped relationship.

3.2. Mediating Mechanism

Under the intelligent manufacturing mode, manufacturers’ trend of servitization transformation is apparent [8]. Manufacturers’ response speed to customers’ personalized and differentiated needs has improved dramatically [34]. Although manufacturing companies can obtain improved economic benefits when they carry out internal servitization, they will face various organizational management risks. Especially in the integration of services related to R & D and design, it is difficult to give full play to the synergistic effect between manufacturing and service within the enterprise [35,36]. Large-scale manufacturing enterprises have strong internal integration capabilities for service. However, they also need to rely on external service providers, especially the knowledge-intensive producer service sector, to ensure the smooth progress of enterprise service and increase the added value of enterprise products [37]. In a word, servitization will increase the demand for producer services. In addition, when manufacturing intelligence helps companies move towards both ends of the value chain, manufacturers’ market demand for producer services will become more diversified. As the diversified producer service environment can provide rich and diverse business support, it can effectively integrate scientific research, production, operation, and other capabilities in the supply chain and brand reputation. Therefore, manufacturers can fully integrate and utilize various resource elements to avoid and diversify risks. Then, the service needs of different stages in the upgrading of the manufacturing industry can be effectively met. Moreover, diversified market demands enable producer services to obtain more significant economies of scope and scale benefits, thus driving the diversified agglomeration of producer services [38,39]. To summarize, manufacturing intelligentization can promote the diversified agglomeration of producer services through the demand for diversified services.

However, with manufacturers’ further increase of intelligent investment, the supply chain platform services based on intelligent collaboration will gradually improve [40]. Furthermore, the spatial division of labor, such as research and development, procurement, production, transportation, and storage, will be more systematic and standardized. In this case, the space–time distance of manufacturing will be continuously compressed by the existence of the intelligent supply chain. Transportation and communication costs between regions will decrease to an extremely low state. When the critical value is exceeded, the dynamic regional industrial transfer process will occur. The same industry eventually agglomerates in the most advantageous spatial location [41]. The traditional producer service comprehensive agglomeration area will gradually transform into a specialized service agglomeration, forming with a specific service as the primary function. In other words, the original diversified agglomeration mode changes under the trend of high intelligence. Agglomeration areas dominated by a single producer service will emerge, such as financial service agglomeration, technical service agglomeration, and logistics agglomeration [42]. Therefore, when manufacturers’ intelligent investment is at a high stage, its driving effect on the diversified agglomeration of producer services will be weakened and then inhibited. Based on the above analysis, this study proposes hypothesis H2:

H2. 

Manufacturing intelligentization has an inverted U-shaped influence on the diversified agglomeration of producer services.

From the perspective of knowledge spillover, the agglomeration of different producer services can provide more possibilities for “face-to-face” communication between different producer services and between producer services and local innovation subjects. Therefore, diversified producer services can provide more transfer channels for complementary innovative knowledge with complexity, ambiguity, and uncertainty. From the perspective of an innovation environment, a diversified productive service environment can form an innovation network between the supply and demand side of the innovation chain, especially between productive service providers and manufacturers. Innovators can reduce transaction costs and promote the effective dissemination of information and technology in the region through the network environment. Further, when information is limited, expensive, and in the presence of opportunistic behavioral events, the spatial and relational proximity brought by the innovative network can facilitate coordinated decision making. The network environment resists opportunistic behaviors that endanger collective values by cultivating and forming trust mechanisms and social sanctions between individuals in the agglomeration area to reduce the difficulty of coordination in the innovation process [43]. Based on this, the article makes the following assumption:

H3. 

Diversified agglomeration of producer services can improve innovation performance and play a nonlinear mediating role in the impact of manufacturing intelligentization on innovation performance.

4. Methods

4.1. Variable Description

4.1.1. Dependent Variable

Innovation performance The article uses the logarithmic form of new product sales revenue to measure. On the one hand, new product sales revenue can reflect the commercialization level of R & D achievements; on the other hand, it can also reflect the final economic value of other R & D effects such as the improvement of technological processes and product quality. These are aspects of innovation that are difficult to measure using data on patent applications or grants.

4.1.2. Independent Variable

Manufacturing intelligentization: Drawing on the definition of intelligent manufacturing from Li (2020), the study constructs a comprehensive intelligentization index of manufacturing from three aspects: intelligent technology, intelligent application, and intelligent benefit [44]. Intelligent technology reflects the physical basis of manufacturing intelligentization and is the premise and guarantee of intelligentization, and the investment in fixed assets of intelligent facilities is the source of power for the progress of intelligent technology. Thus, the article uses fixed asset investment in information transmission, computer services, and software industries as proxy indicators for intelligent technology. Intelligent application is the guarantee and efficiency of manufacturing intelligent technology, reflecting the technological development and application in intelligentization. The study uses the number of patent applications in the manufacturing of electronic and communication equipment to measure this indicator. The intelligent benefit reflects the market profitability in manufacturing intelligentization. The study uses the profit from manufacturing electronic and communication equipment to measure this.

The construction of indicators of manufacturing intelligentization are shown in

Table 1. Construction of manufacturing intelligentization.

Categories	Indicators
Intelligent technology	Fixed asset investment in information transmission, computer services, and software industries
Intelligent application	Number of patent applications for electronic and communication equipment manufacturing industry
Intelligent benefit	Total profit of electronic and communication equipment manufacturing

. Moreover, the study uses the entropy weight method to weigh the indicators.

4.1.3. Mediating Variable

Diversified agglomeration of producer services is expressed by the improved Combes (2000) [45] industry diversification index. This study classifies “Transportation, warehousing and postal industry”, “Information transmission, computer service and software industry”, “Wholesale and retail trade industry”, “Financial industry”, “Leasing and commercial service industry”, “Scientific research and technical service industry” and “Water conservancy, environment and utility management” as producer services [46]. The index is established as shown in Equation (1) and calculated yearly. $D{V}_i$ is the level of diversified agglomeration of producer services in region i; $E_{i{s}^{\prime }}$ is the employment of a sector s′ of producer services in region i, except sector s; $E_i$ is the number of people employed in region i; $E_{is}$ is the number of employees in a sector s of producer services in region i; $E_{s^{\prime }}$ is the number of employed persons in a sector s’ of producer services, except the sector s at the national level; $E_s$ is the employment in the producer service sector s in the country; $E$ is the total employment in the country. The larger the $D{V}_i$, the higher the diversified agglomeration level of producer services.

$$D{V}_i={\displaystyle \sum _s\frac{E_{is}}{E_i}\left[\frac{1/{\displaystyle \sum _{s^{\prime }=1.s^{\prime }\ne s}^n{\left[E_{i{s}^{\prime }}/(E_i-E_{is})\right]}^2}}{1/{\displaystyle \sum _{s^{\prime }=1.s^{\prime }\ne s}^n{\left[E_{s^{\prime }}/(E-E_s)\right]}^2}}\right]}$$

(1)

4.1.4. Control Variables

This study incorporates human capital (HC), fixed capital stock (Fixcap), government expenditure (Gov), technology market activity (CTR), population density (Pdensity), and cost of living (Lifecost) into the model to capture the different levels of economic development in the region.

The control variables were calculated as follows: human capital level (HC), measured by the proportion of students in regular institutions of higher education to the regional population; fixed capital stock (Fixcap), calculated using the perpetual inventory method and divided by real GDP (deflated at comparable prices in 2000); government expenditure (Gov), measured by the logarithm of regional government fiscal expenditure; technology market activity (CTR), measured by the logarithm of the technology contract turnover; population density (Pdensity), measured by the number of people per square kilometer in the region; the cost of living (Lifecost) is the ratio of the per capita consumption expenditure to the per capita disposable income.

The article’s data are the panel data of 30 provinces in China (excluding Tibet, Hong Kong, Macao, and Taiwan) from 2008 to 2020. All data involved in this article come from the State Statistical Bureau, China Statistical Yearbook 2008–2020 repository, and the EPS China Data platform. The names of all variables and their descriptive statistics are shown in

Table 2. List of variables.

Variable Type	Variable Name	Variable Symbol
Dependent Variable	Innovation performance	Innov
Independent Variable	Manufacturing intelligentization	Intel
Mediating Variable	Diversified agglomeration of producer services index	DV
Control Variables	Human capital level	HC
	Fixed capital stock	Fixcap
	Government expenditure	Gov
	Technology market activity	CTR
	Population density	Pdensity
	The cost of living	Lifecost

and

Table 3. Descriptive statistics results.

Variable	Obs	Mean	Std. Dev.	Min	Max
Innov	390	7.518	1.590	2.148	10.699
Intel	390	0.065	0.109	0.004	0.972
DV	390	0.219	0.051	0.077	0.348
HC	390	0.026	0.009	0.010	0.068
Fixcap	390	10.758	5.792	3.819	36.749
Gov	390	17.376	0.702	14.993	18.976
CTR	390	13.626	1.843	8.623	17.962
Pdensity	390	0.048	0.073	0.001	0.429
Lifecost	390	0.580	0.245	0.184	1.296

.

4.2. Model Design

According to theoretical analysis and research assumptions, this study mainly examines the nonlinear impact of manufacturing intelligentization on innovation performance and its internal mechanism in the empirical part. Drawing on Edwards’ (2007) [47] moderated mediation mechanism test method, the following individual-fixed and time-fixed panel model is constructed for analysis. This article first analyzes the impact of intelligentization on innovation performance, as shown in model (2). SqIntel is the quadratic term of Intel. Then, the next step is to examine the mediating mechanism of the diversified agglomeration of producer services, as shown in models (3) and (4). The criterion is: if the coefficient $b_{2,2}$ in model (3) is in the same direction as the coefficient $b_{1,2}$ in model (2), and the intermediary variable coefficient $b_{3,3}$ in model (4) is significant, it can be determined that the impact of intelligentization on innovation performance has a nonlinear transmission mechanism of diversified agglomeration of producer services.

$$Inno{v}_{it}=b_{1,0}+b_{1,1}Inte{l}_{it}+b_{1,2}SqInte{l}_{it}+{\displaystyle \sum _{j=1}^6{\delta }_{1,j}{X}_{j,it}}+u_i+v_t+e_{it}$$

(2)

$$D{V}_{it}=b_{2,0}+b_{2,1}Inte{l}_{it}+b_{2,2}SqInte{l}_{it}+{\displaystyle \sum _{j=1}^6{\delta }_{2,j}{X}_{j,it}}+u_i+v_t+e_{it}$$

(3)

$$Inno{v}_{it}=b_{3,0}+b_{3,1}InIte{l}_{it}+b_{3,2}SqInte{l}_{it}+b_{3,3}D{V}_{it}+b_{3,4}Inte{l}_{it}\times D{V}_{it}+{\displaystyle \sum _{j=1}^6{\delta }_{3,j}{X}_{j,it}}+u_i+v_t+e_{it}$$

(4)

4.3. Regression Results

From the significant directions of the coefficients of Intel and SqIntel in columns (1) and (2) of

Table 4. Benchmark regression results.

	(1)	(2)	(3)	(4)
	Innov	Innov	DV	Innov
Intel	3.182 ***	3.710 ***	0.238 ***	6.918 **
	(3.542)	(3.910)	(5.534)	(2.523)
SqIntel	−1.959 **	−2.439 ***	−0.217 ***	−2.198 **
	(−2.471)	(−3.023)	(−5.932)	(−2.368)
DV				5.183 ***
				(3.981)
Intel × DV				−13.513
				(−1.646)
HC		33.069 ***	2.616 ***	23.268 **
		(3.216)	(5.606)	(2.202)
Fixcap		0.023 ***	0.000	0.021 ***
		(2.894)	(0.721)	(2.715)
Gov		0.359	0.016	0.275
		(1.514)	(1.524)	(1.179)
CTR		−0.047	−0.000	−0.050
		(−1.290)	(−0.030)	(−1.397)
Pdensity		1.920	0.141	0.696
		(0.583)	(0.942)	(0.214)
Lifecost		−0.276	0.021	−0.389
		(−0.853)	(1.459)	(−1.223)
_cons	7.344 ***	0.703	−0.166	1.455
	(148.512)	(0.173)	(−0.901)	(0.365)
Province FE	yes	yes	yes	yes
Time FE	yes	yes	yes	yes
N	390	390	390	390
R2	0.962	0.965	0.931	0.967

Note: **, and *** indicate significance levels p < 0.05, and p < 0.01, respectively.

, it can be seen that the impact of manufacturing intelligentization on innovation performance presents a significant inverted “U”-shaped relationship. A lower level of intelligentization can promote innovation performance, but when it exceeds a certain threshold, intelligentization will inhibit the development of regional innovation achievements. H1 is validated. The significant coefficients of Intel and SqIntel in column (3) show that the impact of intelligentization on the diversified agglomeration of producer services also presents a nonlinear relationship of “promoting first and then inhibiting”. H2 is validated. Observing column (4) in Table 4, the influence coefficient of DV on innovation performance is significantly positive, indicating the intelligentization of the manufacturing industry has an inverted “U”-shaped impact on innovation performance through the channel of diversified agglomeration of producer services. H3 is validated.

4.4. Robustness Tests

This paper conducts the following robustness tests: (1) principal component method is used to recalculate the index of manufacturing intelligentization; (2) the logarithmic R & D expenditure is included in the models (2)~(4) as a control variable; (3) drawing from Duranton and Puga’s (2001) [48] measure of diversified agglomeration, this study recalculates the diversified agglomeration index of producer services.

The robustness test results are shown in

Table 5. Results of robustness tests.

	Principal Component			Include R & D			Change DV’s Measure
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	Innov	DV	Innov	Innov	DV	Innov	DV	Innov
Intel	0.324 ***	0.023 ***	0.588 **	2.699 ***	0.244 ***	5.750 **	0.028 ***	21.565
	(3.519)	(5.395)	(2.120)	(2.813)	(5.473)	(2.139)	(5.870)	(0.234)
Sqintel	−0.027 **	−0.003 ***	−0.021	−1.533 *	−0.222 ***	−1.227	−0.022 ***	−2.045 *
	(−2.330)	(−5.618)	(−1.624)	(−1.873)	(−5.843)	(−1.316)	(−5.433)	(−1.894)
DV			4.482 ***			5.294 ***		25.475 **
			(3.806)			(4.169)		(2.054)
Intel × DV			−1.197			−13.182		−18.579
			(−1.386)			(−1.646)		(−0.201)
HC	32.595 ***	2.536 ***	22.895 **	20.172 *	2.690 ***	9.714	0.288 ***	26.707 **
	(3.147)	(5.402)	(2.155)	(1.916)	(5.496)	(0.902)	(5.601)	(2.399)
Fixcap	0.021 ***	0.000	0.019 **	0.029 ***	0.000	0.027 ***	0.000 ***	0.021 **
	(2.637)	(0.480)	(2.466)	(3.686)	(0.612)	(3.544)	(2.698)	(2.527)
Gov	0.350	0.015	0.264	−0.077	0.019	−0.171	0.001	0.335
	(1.470)	(1.388)	(1.123)	(−0.301)	(1.594)	(−0.686)	(0.729)	(1.419)
CTR	−0.048	−0.000	−0.050	−0.019	−0.000	−0.021	−0.000	−0.047
	(−1.319)	(−0.164)	(−1.390)	(−0.521)	(−0.124)	(−0.596)	(−0.074)	(−1.300)
Pdensity	1.870	0.126	0.909	0.565	0.149	−0.691	−0.078 ***	3.746
	(0.565)	(0.841)	(0.279)	(0.175)	(0.988)	(−0.217)	(−4.744)	(1.101)
Lifecost	−0.291	0.021	−0.405	−0.290	0.021	−0.406	0.004 ***	−0.386
	(−0.896)	(1.445)	(−1.265)	(−0.918)	(1.463)	(−1.308)	(2.721)	(−1.184)
RD				0.398 ***	−0.002	0.406 ***
				(4.116)	(−0.504)	(4.296)
_cons	1.137	−0.121	2.070	2.674	−0.177	3.489	0.967 ***	−23.932 *
	(0.278)	(−0.651)	(0.514)	(0.669)	(−0.954)	(0.892)	(47.634)	(−1.891)
Province FE	yes	yes	yes	yes	yes	yes	yes	yes
Time FE	yes	yes	yes	yes	yes	yes	yes	yes
N	390	390	390	390	390	390	390	390
R2	0.965	0.931	0.966	0.967	0.931	0.968	0.965	0.966

Note: *, **, and *** indicate significance levels p < 0.1, p < 0.05, and p < 0.01, respectively.

, where the Intel and SqIntel coefficients of columns (1), (2), (4), (5), and (7) are all significant at the 10% level, and the DV coefficients of columns (3), (6), and (8) are all significant at the 5% level. Additionally, the significant directions are the same as those in Table 4. Thus, the empirical results of this article are relatively stable.

4.5. Test of the Moderating Effect

The above research results show that manufacturing intelligentization has an inverted “U”-shaped influence relationship on innovation performance. This result shows that as the intelligence level of the manufacturing industry improves, it has a diminishing marginal effect of “promoting first and then inhibiting” on the innovation performance. Additionally, this influence exists in the channel of diversified agglomeration of producer services. To break the constraint of diminishing marginal effect of the impact of intelligentization on innovation performance, the study has the following aims:

Considering that when the level of human capital in the region is at a high stage, its workforce skill structure will be gradually optimized and upgraded. In this case, more skilled labor will be matched with intelligent development, thereby enhancing enterprises’ adaptability and knowledge absorption rate in the face of complex, intelligent manufacturing systems. In addition, highly skilled laborers have a stronger ability to discover and solve problems. In the intelligent manufacturing environment, they can make more targeted use of intelligent technologies and find practical evidence applicable to new problems in massive data, to propose better solutions. Therefore, improving human capital can strengthen the positive impact of manufacturing intelligentization on innovation performance to a certain extent. In addition, higher human capital often carries more tacit knowledge. When enterprises in the agglomeration area consider the transformation from agglomeration to decentralized development, the cost of tacit knowledge transfer will gradually become a restraining factor for enterprises to transfer across regions. Therefore, the following model is further constructed to study the moderating influence of human capital on the relationship between intelligentization and innovation performance.

$$\begin{array}{ll}Inno{v}_{it}& ={\beta }_{1,0}+{\beta }_{1,1}Inte{l}_{it}+{\beta }_{1,2}SqInte{l}_{it}+{\beta }_{1,3}Inte{l}_{it}\times H{C}_{it}+{\beta }_{1,4}SqInte{l}_{it}\times H{C}_{it}\\ & +{\beta }_{1,5}H{C}_{it}+{\displaystyle \sum _{j=1}^6{\theta }_{1,j}{X}_{j,it}}+u_i+v_t+e_{it}\end{array}$$

(5)

$$\begin{array}{ll}D{V}_{it}& ={\beta }_{2,0}+{\beta }_{2,1}Inte{l}_{it}+{\beta }_{2,2}SqInte{l}_{it}+{\beta }_{2,3}Inte{l}_{it}\times H{C}_{it}+{\beta }_{2,4}SqInte{l}_{it}\times H{C}_{it}\\ & +{\beta }_{2,5}H{C}_{it}+{\displaystyle \sum _{j=1}^6{\theta }_{2,j}{X}_{j,it}}+u_i+v_t+e_{it}\end{array}$$

(6)

$$\begin{array}{ll}Inno{v}_{it}& ={\beta }_{3,0}+{\beta }_{3,1}Inte{l}_{it}+{\beta }_{3,2}SqInte{l}_{it}+{\beta }_{3,3}Inte{l}_{it}\times H{C}_{it}+{\beta }_{3,4}SqInte{l}_{it}\times H{C}_{it}\\ & +{\beta }_{3,5}D{V}_{it}+{\beta }_{3,6}Inte{l}_{it}\times D{V}_{it}+{\beta }_{3,7}H{C}_{it}+{\displaystyle \sum _j^5{\theta }_{3,j}{X}_{j,it}}+u_i+v_t+e_{it}\end{array}$$

(7)

Based on models (2)~(4), models (5)~(7) are constructed to include the interaction terms between HC and Intel, as well as the interaction terms between HC and SqIntel. The verification results are shown in

Table 6. Moderating effect of human capital.

	(1)	(2)	(3)	(4)
	Innov	Innov	DV	Innov
Intel	14.191 ***	13.628 ***	0.069	16.241 ***
	(3.877)	(3.742)	(0.411)	(3.832)
SqIntel	−12.997 ***	−12.919 ***	−0.160	−13.160 ***
	(−2.851)	(−2.851)	(−0.773)	(−2.922)
Intel × HC	−374.447 ***	−331.704 ***	6.117	−338.093 ***
	(−3.196)	(−2.819)	(1.137)	(−2.899)
SqIntel × HC	397.120 ***	373.824 **	−2.682	394.877 ***
	(2.593)	(2.450)	(−0.384)	(2.639)
DV				5.200 ***
				(4.019)
Intel × DV				−11.005
				(−1.276)
HC	31.276 ***	34.535 ***	2.561 ***	23.928 **
	(4.173)	(3.379)	(5.477)	(2.286)
Fixcap		0.020 **	0.000	0.018 **
		(2.558)	(0.929)	(2.394)
Gov		0.335	0.018	0.253
		(1.423)	(1.633)	(1.092)
CTR		−0.042	−0.000	−0.044
		(−1.154)	(−0.167)	(−1.249)
Pdensity		2.081	0.126	0.911
		(0.636)	(0.843)	(0.281)
Lifecost		−0.305	0.022	−0.419
		(−0.951)	(1.481)	(−1.330)
_cons	6.502 ***	0.998	−0.182	1.743
	(33.233)	(0.248)	(−0.989)	(0.441)
Province FE	yes	yes	yes	yes
Time FE	yes	yes	yes	yes
N	390	390	390	390
R2	0.965	0.966	0.932	0.967

Note: **, and *** indicate significance levels p < 0.05 and p < 0.01, respectively.

. It can be seen from columns (2) and (4) that the significant directions of Intel and SqIntel coefficients are consistent with columns (2) and (4) in Table 4. After adding the interaction item, the inverted “U”-shaped impact of manufacturing intelligentization on innovation performance still exists. From the significant direction of the interaction coefficients of Intel and SqIntel with HC in columns (2) and (4) of Table 6, it can be seen that HC can inversely regulate the inverted “U”-shaped impact of intelligentization on innovation performance. To more intuitively show the relationship between Intel, HC, and Innov, the article draws a three-dimensional relationship diagram between the three based on the variable coefficients in column (2) of Table 6. According to Figure 1, when the marginal effect of Intel diminishes to the point of inhibiting innovation performance, appropriately raising the level of human capital can alleviate the inhibitory effect. In other words, paying attention to the coordinated development of manufacturing intelligence and human capital can alleviate the squeeze of innovation performance suffered by the manufacturing industry in a high-intelligence environment. The synergistic development of the two can further promote innovation and sustainable economic development.

5. Discussion

The arrival of the era of intelligent manufacturing marks the deep integration of informatization and industrialization and the beginning of manufacturing enterprises to move towards high-end value chains and service-oriented manufacturing. Manufacturing intelligentization can be a driving force for economic innovation and sustainable development. Nevertheless, from the empirical data, Chinese manufacturing companies have not fully realized the potential of intelligence in manufacturing to empower economic innovation and development. Although the relatively low intelligence of the manufacturing industry can improve China’s regional innovation performance, when the regional manufacturing industry overinvests in intelligent transformation, innovation will be squeezed. At the same time, in the process of service-oriented transformation of the manufacturing industry, the impact of manufacturing intelligentization on innovation performance of “promoting first and then inhibiting” has an intermediary channel of diversified agglomeration of producer services. This mechanism more clearly shows the action law and operation logic of intelligent manufacturing strategy on innovation, which is helpful to guide social practice better. In addition, the marginal diminishing impact of intelligentization on innovation can be significantly alleviated by optimizing human capital structure. To summarize, focusing on managing intelligent investment, the basic environment of producer services, and human capital structure are significant to the innovative and sustainable development driven by manufacturing intelligence.

5.1. Practical Implications

First of all, in the current environment where the manufacturing value of products is gradually decreasing and the economy has difficulties in sustainable development, Chinese manufacturers need to rely on intelligence to improve innovation performance. On the one hand, manufacturers can obtain a large amount of multi-source heterogeneous data by introducing intelligent technology, building physical information systems, and linking the entire product life cycle to build a “product–service” solution pattern. Through big-data analysis and prediction, a large amount of explicit knowledge is obtained, and the product development cycle is shortened to promote product innovation, value increase, and sustainable development. On the other hand, manufacturers have a threshold of investing in intelligence. Through physical information systems and service transformation, enterprises can link more physical equipment, manufacturers, and service providers in the supply chain, to obtain more explicit knowledge, form innovation potential, and improve product value and market competitiveness. However, due to the limited ability of people to obtain and analyze information, the excessive interconnection and concentration of information in high intelligence will lead to information overload and information barriers, raise the cost of innovation, and cause a technology lock. Therefore, manufacturers need to cultivate certain intelligent investment management capabilities, improve the innovation driving force of intelligent technology, and strengthen innovation results.

Secondly, manufacturers should pay attention to the match between intelligent investment in manufacturing and producer service environment. The influence of manufacturing intelligence on innovation performance has the internal mechanism of producer services’ diverse agglomeration. Manufacturers have a service-oriented trend when investing in intelligent technology, which shows that enterprises inject more and more productive service elements into the production process to extend the value of products. The service input caused by intelligence will attract diversified producer service enterprises to gather in intelligent areas, to adapt to the service needs of manufacturers in different stages of service transformation. However, the highly intelligent production environment will compress transportation costs and transaction costs to a minimum, thus inducing the producer service providers to transfer to the area with the highest factor endowment, which is not conducive to the diversification and agglomeration of service providers, thus inhibiting the innovation brought by diversification. Therefore, when manufacturers drive innovation development through intelligence, they should pay attention to the matching degree between the environment of producer service providers and the level of intelligence, to effectively enhance innovation output.

For governments and policymakers, it is necessary to strengthen the in-depth application of information technology in manufacturing and guide qualified small and medium-sized manufacturing enterprises to carry out the digital and intelligent transformation, to avoid duplication or excessive investment. The infrastructure of various types of productive services should be further improved. A market business environment with fair competition and a transparent legal environment should be established. These will provide basic conditions for the agglomeration of producer services, expand market size, and cultivate a variety of producer service suppliers. Furthermore, it is necessary to improve the adaptability between the skill structure of labor supply and the intelligence level of manufacturing through the policy system. Further, the coordinated development of intelligent manufacturing and higher education should be promoted by improving the popularity and quality of higher education.

5.2. Theoretical Implications

Firstly, this study reveals that manufacturing intelligence is a key nonlinear factor affecting innovation performance growth, which enriches the existing research framework on the relationship between intelligent technology and innovation. Existing literature mainly studies the positive relationship between intelligent technology and innovation from a linear perspective [4,5], without paying attention to the potential negative impact of excessive investment in intelligent technology on product innovation and new knowledge output. This paper complements related research from a nonlinear perspective and verifies the inverted U-shaped relationship between manufacturing intelligence and innovation performance.

Secondly, the diversified agglomeration of producer services was taken as the key intermediary factor, enriching the research on intelligence’s influence on innovation performance. The existing literature mainly focuses on the linear direct impact of intelligence on innovation [4] and studies the intermediary mechanism from the perspective of innovation ability [5], while there are few studies on the diversification and agglomeration of producer services as a nonlinear potential mechanism. Under the background of economic service, this paper reveals the nonlinear effect path of intelligence on innovation performance from the perspective of diversification and agglomeration of producer services. It not only expands the research on the nonlinear influence channels of intelligence on innovation but also enriches the literature on innovation development under the trend of servitization.

Thirdly, to alleviate the marginal diminishing constraint of the impact of manufacturing intelligence on innovation performance, this paper uses the level of human capital to verify its moderating effect on the inverted U-shaped impact from the perspective of labor structure optimization. To some extent, this reveals how to promote the sustainable development of innovation strategy driven by intelligence.

6. Conclusions

In the era of intelligent manufacturing, intelligent technology will be embedded in all links of the innovation chain to promote innovation and sustainable development in the manufacturing industry. The article mainly analyzes the nonlinear impact of manufacturing intelligentization on innovation performance and examines the mediating role of the diversified agglomeration of producer services. The study found that the impact of manufacturing intelligentization on innovation performance presents a significant inverted “U”-shaped relationship; at the same time, this impact has a nonlinear channel of diversified agglomeration of producer services. Further tests point out that, to a certain extent, the improvement of the human capital level can restrain the diminishing marginal effect of intelligentization on innovation. That is, it can alleviate the negative effect of higher intelligence levels on regional innovation.

This study has some limitations. First, this study only used the individual and time fixed-effects model for testing and did not consider the interaction between regions. Second, due to the COVID-19 epidemic in 2020, China’s economic development was affected to a certain extent, and the article included the data for 2020 in the study without corresponding consideration. Therefore, the reliability of empirical predictions in the article may be reduced.

For future research, intelligent manufacturing is undoubtedly a means to improve the value of China’s manufacturing industry chain and a significant driving force for China’s economic innovation and sustainable development. However, in the process of servitization, intelligent manufacturers may face a contradiction between innovation and efficiency [36]. For example, intelligent enterprises can quickly respond to and customize the needs of the consumer market. So, should the technology development within an intelligent enterprise be customer-oriented, or should it be based on engineering thinking? Customer orientation is necessary when customizing solutions and advanced services for a client’s business, but engineering thinking is critical to maintaining a culture that supports the development of highly innovative products and solutions. Customized solutions have low reusability and production efficiency of related products, which is different from efficient incremental innovation under engineering thinking. In other words, manufacturers may face the paradox of customized innovation and production efficiency in intelligentization. This paradox is also a pivotal point to be considered in future research on the relationship between intelligence and innovation.