Total factor carbon emission performance, based on production theory, is measured by considering multiple factor inputs, evaluating the combined effects of economic output and carbon dioxide emissions. Digital infrastructure, as a fundamental carrier of data elements, can alter existing production factor configurations through the application of data elements, thus influencing expected economic outputs and unintended carbon dioxide emissions, ultimately affecting urban total factor carbon emission performance.

Due to the significant reliance of digital infrastructure on energy, which generates considerable carbon emissions, the impact of digital infrastructure on urban carbon performance is complex. On one hand, urban digitalization and intelligentization increase information accessibility, leading to efficient resource allocation and improved energy utilization. When energy efficiency improves, carbon emissions per unit of energy usage remain constant, but the generated economic increment increases, enhancing total factor carbon emission performance. On the other hand, considering the energy rebound effect, digital infrastructure might have a counterproductive impact on urban carbon performance. First, the construction of digital infrastructure itself might increase carbon emissions due to its high energy consumption during construction and operation (Avom et al., 2020). Secondly, while digital infrastructure promotes economic growth, current economic and social energy consumption mainly relies on fossil fuels. Economic growth resulting in expanded output scales and technological advancements can lead to increased energy consumption and corresponding carbon emissions (Khazzoom, 1980).

From a macro perspective, digital infrastructure construction promotes clean energy substitution for traditional energy and the replacement of material inputs with data elements, making urban activities cleaner and more efficient by mitigating factor misallocation (Ge et al., 2022). This can effectively reduce carbon emissions while fostering economic growth, thus improving total factor carbon emission performance. However, digital infrastructure requires substantial funding, and given limited resources, the initial investment costs, maintenance costs, and training costs can squeeze investments in energy-saving and carbon-reducing initiatives. The cost effect of digital infrastructure construction could hinder improvements in carbon performance (Salahuddin and Alam, 2015). On a micro level, the introduction of data elements alters enterprise production scales, and digital platforms influence how enterprises and consumers access information, affecting the traditional supply-demand chain of products in the market. The supply of green products and the green consumption chain emerging from this can further impact urban total factor carbon performance.

In summary, digital infrastructure can enhance urban carbon performance by improving energy efficiency, utilizing clean energy to replace fossil fuels, and exhibiting the energy rebound and cost effects that suppress improvements in carbon performance. Thus, the impact of digital infrastructure construction on urban carbon performance exhibits a complex bidirectional relationship. Based on this, the following competing hypotheses are proposed.

Hypothesis 1a

Digital infrastructure positively impacts urban total factor carbon emission performance.

Hypothesis 1b

Digital infrastructure negatively impacts urban total factor carbon emission performance.

For consumers, digital infrastructure construction can reduce carbon emissions in household activities by altering lifestyles and consumption habits, thereby enhancing total factor carbon emission performance. The rapid advancement of digitalization promotes sustainable green consumption (Peng Y. et al., 2024). Digital infrastructure construction makes life more convenient, promoting paperless learning and working methods. Internet platforms enable consumers to complete shopping, medical consultations, and other activities online, encouraging the development of second-hand markets and shared services. In 2023, the national online retail sales of physical goods reached 13,017.4 billion yuan, reflecting a notable shift from offline to online consumption, a hallmark of modern consumption modes (Ru and Deng, 2023). This convenience and resource recycling reduce consumer energy demand, particularly lowering energy use during travel, and subsequently cutting carbon emissions.

Moreover, internet applications and environmental public welfare campaigns significantly influence consumer consumption concepts, enhancing low-carbon consciousness through virtual low-carbon behaviors that translate into real-world low-carbon actions. For example, participants in the “Ant Forest” program are more likely to choose walking or public transportation over private cars to earn “energy points.” Similarly, influenced by environmental awareness, consumers may prefer clean fuels when selecting energy sources, promoting low-carbon energy consumption patterns (Ma, 2020). When purchasing vehicles, consumers may increasingly prefer new energy vehicles, thus reducing the use of chemical energy and significantly lowering carbon emissions, enhancing total factor carbon emission performance. Based on this, the following hypothesis is proposed.

Hypothesis 2

Digital infrastructure construction influences total factor carbon emission performance by altering consumer consumption patterns.

Digital infrastructure construction can transform enterprise production modes by facilitating corporate transformation and upgrading, affecting total factor carbon emission performance during production. In enterprise production, digital infrastructure construction empowers businesses to optimize resource allocation, enhance production factor efficiency, reduce unit energy consumption, and improve capacity utilization rates (Fu et al., 2024). Furthermore, digital facilities enhance enterprise information systems, prompting companies to establish databases and new digital systems that simulate production and emissions processes to improve efficiency. The lowered information acquisition threshold strengthens inter-enterprise cooperation, diversifies resources for green technology innovation, and enhances overall human capital, fostering green technology advancements driven by talent. This contributes to upgrading production processes, reducing costs, and encouraging investments in energy-saving and carbon-reducing efforts, ultimately promoting green production and product innovation.

Additionally, digital infrastructure has led to the emergence of internet-based online revenue businesses, which greatly substitute data elements for traditional energy, thereby driving economic development while significantly reducing energy consumption and carbon emissions. The appearance of these low-carbon enterprises fosters competition, influencing the flow and allocation efficiency of production factors, eliminating high-energy-consuming firms, and reducing the proportion of highly polluting enterprises. This leads to a decrease in energy consumption and carbon emissions per unit of output, ultimately enhancing carbon performance (Wang et al., 2024). Simultaneously, reallocating resources promotes investment in low-carbon technological innovation, contributing to low-carbon industrial development (Ang, 1999). Based on this, the following hypothesis is proposed:

Hypothesis 3

Digital infrastructure construction influences total factor carbon emission performance by altering enterprise production.

Based on the above analysis and hypotheses, the article proposes the theoretical framework diagram in Figure 1.

Based on the approved list of cities involved in the “Broadband China” strategy by the Ministry of Industry and Information Technology and the National Development and Reform Commission of China, 82 cities implementing the strategy were selected as the treatment group, while the remaining cities that did not implement the strategy formed the control group. To examine whether the “Broadband China” strategy improved the total factor carbon emission performance of pilot cities, a gradual Difference-in-Differences (DID) model was constructed as the Equation 1: U E I i t = α 0 + α 1 D I D + γ C o n t r o l i t + μ i t + φ t + ε i t (1)

Where: i and t represent cities and years; The dependent variable ( U E I ) represents “total factor carbon emission performance”; The core explanatory variable ( D I D ) is a dummy variable for the “Broadband China” pilot cities; C o n t r o l variables include city-level indicators such as “government size,” “city size,” and “economic development level”; μ i indicates city fixed effects, φ t represents time fixed effects, and ε i t is the random disturbance term.

Due to data limitations on residential electricity consumption, which are only available until 2019, and the restricted number of cities with matching corporate data, this study uses panel data from 185 prefecture-level cities in China from 2010 to 2019, amounting to 1,850 observations. Data sources include the China Urban Statistical Yearbook, China Energy Statistical Yearbook, China Statistical Yearbook, the National Geophysical Data Center (NGDC), the China Emissions Accounts Database (CEADs), CSMAR, the National Bureau of Statistics, various provincial statistical yearbooks, and municipal statistical yearbooks. Missing data were supplemented using interpolation methods. Descriptive statistics of key variables are presented in Table 1.

The baseline regression results of the impact of digital infrastructure on urban total factor carbon emission performance are shown in Table 2. Under the control of city and year fixed effects, control variables were gradually added to the regression. From the regression results in each column, it can be observed that the estimated coefficients of the core explanatory variable D I D are significantly positive at the 5% level, indicating that cities implementing the “Broadband China” strategy significantly enhance their total factor carbon emission performance. Specifically, from the estimated coefficient in column (5), it can be observed that after adding all control variables, the core explanatory variable D I D is 0.0228, which is significant at the 1% level, implying that the “Broadband China” strategy increases the total factor carbon emission performance of pilot cities by 2.28% compared to non-pilot cities.

This verifies Hypothesis 1a: Digital infrastructure positively impacts urban total factor carbon emission performance.

Using the DID model requires satisfying the parallel trend assumption, which states that the treatment and control groups should exhibit parallel changes before policy implementation. Following Beck et al. (2010), the parallel trend test was conducted. The results are illustrated in Figure 2, using the year before the implementation of the “Broadband China” strategy as the base period. Before the policy implementation, there was no significant difference in total factor carbon emission performance between pilot and non-pilot cities. The differences became significant 2 years after policy implementation. Due to the time needed for digital infrastructure construction, application, and the influence on market participants like consumers and enterprises, short-term differences may not be obvious immediately after the policy announcement, indicating a lag effect in policy impact. This observation meets the parallel trend assumption, validating the use of the DID method.

To address potential unobserved variables at the city-year level that may confound policy evaluation effects, a placebo test was conducted. A randomly generated experimental group list for “Broadband China” pilots was created, and the baseline model was repeated 500 times. As shown in Figure 3, the kernel density distribution of estimated coefficients and their P-values approximately follow a normal distribution, suggesting robustness in the regression results.

The baseline regression results suggest that digital infrastructure construction, represented by the “Broadband China” strategy, can enhance urban total factor carbon emission performance. The robustness of these findings was further verified through various methods, including changes in model specifications, adjustments to policy timelines, exclusion of other policy interferences, removal of extreme values, and replacement of core explanatory variables.

To mitigate potential endogeneity issues, the Two-Stage Least Squares (2SLS) method was adopted for testing. Following the approach of Fang et al. (2023), an interaction term was constructed using the distance from each city to Hangzhou and the time trend, and logarithmic processing ( i v ) was applied. This served as an instrumental variable for the “Broadband China” strategy. The results are presented in Table 6. Specifically, the results in column (1) demonstrate the rationality of selecting the instrumental variable. Additionally, cities closer to Hangzhou are more likely to become “Broadband China” pilot cities. In column (2), the coefficient of the core explanatory variable D I D is significantly positive, indicating that even after considering endogeneity, digital infrastructure still enhances urban total factor carbon emission performance.

The above empirical results indicate that digital infrastructure can effectively improve urban total factor carbon emission performance. Based on the previously discussed mechanisms, this section further explores the roles of enterprise production transformation and consumer behavior transformation in the impact of digital infrastructure on carbon performance. Specifically, as the Equations 3, 4, following the approach of Baron and Kenny (1986), this study tests the underlying mechanisms. M i t = α 0 + β D I D + γ C o n t r o l i t + μ i + φ t + ε i t (3) U E I i t = α 0 + β D I D + θ M i t + γ C o n t r o l i t + μ i t + φ t + ε i t (4)

Where: M i t represents the mechanism variables, measured by household electricity consumption ( e l c ) and enterprise green product innovation ( g p ), reflecting green consumer behavior and green products in the market.

The results, shown in Table 7, indicate that digital infrastructure promotes household consumption transformation, thus enhancing total factor carbon emission performance. Digital infrastructure increases residential electricity consumption, and this increase contributes to better carbon performance. This suggests that as digital infrastructure construction advances, residents adopt more digitalized lifestyles, shifting from traditional chemical energy to electricity. Changes in energy consumption structures directly impact total carbon emission performance, showing that the transformation of consumer behavior acts as a partial mediator in the influence of digital infrastructure, supporting Hypothesis 2.

When exploring the mediation effect of enterprise green product innovation, the data was matched further, incorporating firm-level control variables and encompassing 185 cities with 3,203 listed companies from 2010 to 2019, totaling 23,620 data entries. The results, presented in columns (3) and (4) of Table 7, reveal that digital infrastructure fosters green product innovation among enterprises, subsequently improving total factor carbon emission performance. Thus, enterprise green product innovation partially mediates the relationship between digital infrastructure and carbon performance, confirming Hypothesis 3.

In the market, digital infrastructure enhances consumer engagement in green consumption and increases the preference for green products, reducing pollution from consumer activities. Simultaneously, it stimulates enterprise innovation in green products, reducing production-side pollution. There is a reciprocal influence between consumers’ green behaviors and enterprises’ green products. Enterprises’ green products influence consumer purchasing behavior, reshaping consumption patterns. Conversely, consumer demand for green products drives enterprises toward transformation, increasing their investment in and output of green products. Consumer oversight also influences enterprises to adopt greener production methods. Therefore, digital infrastructure indirectly promotes urban total factor carbon emission performance by creating a self-reinforcing cycle of green production and consumption in the market.

To further investigate the potential heterogeneity in the enhancement of urban total factor carbon emission performance by digital infrastructure across various levels, the article explores from the perspectives of consumers and enterprises. Considering that residents’ green consumption behaviors and concepts are related to real economic conditions, the article incorporates the heterogeneity of residents’ income levels for analysis; at the same time, residents’ green consumption concepts and their acceptance and usage of digital infrastructure are acquired by learning, with high quality and level often achieved through learning and training, thus the level of education of residents is taken as another grouping condition for the heterogeneity analysis of consumers. From the perspective of enterprises, the extent to which enterprises accept and use digital infrastructure and their pursuit of green products are often related to corporate management and production technology levels, therefore, enterprise ownership and technological level are taken as grouping conditions for the heterogeneity analysis from the enterprise perspective.

The heterogeneous results of the “Broadband China” strategy on total factor carbon emission performance are shown in Table 8. Based on the disposable income of urban residents, cities are categorized into high-income and low-income groups. Cities with a disposable income above the mean are designated as high-income cities, while those below are considered low-income cities. Through heterogeneity analysis using interaction terms, it can be seen that digital infrastructure is more effective in improving total factor carbon emission performance in high-income cities compared to low-income cities.

Due to data availability, the number of local university students is used as a proxy for residents’ education level. Cities with a student population above the mean are considered high-education cities, and those below are low-education cities. As shown in column (2) of Table 8, digital infrastructure has a more significant impact on improving total factor carbon emission performance in high-education cities compared to low-education cities. This may be because consumers’ application of data elements and their low-carbon behaviors are related to their education levels. Consumers with higher education levels tend to have better low-carbon literacy and engage in more environmentally friendly behaviors.

Therefore, digital infrastructure construction can effectively improve urban total factor carbon emission performance by influencing residents’ consumption behavior transformation, especially when combined with high income and education levels of urban residents. Meanwhile, higher education and income levels of residents can influence their consumption patterns in terms of thinking and behavior, further promoting green development in cities.

From the perspective of businesses, the heterogeneous effects of digital infrastructure on total factor carbon emission performance are shown in columns (3) and (4) of Table 8. Enterprises are classified into state-owned and non-state-owned enterprises, as well as high-tech and non-high-tech industry enterprises, based on ownership and industry. The results indicate that digital infrastructure has a more pronounced effect on promoting total factor carbon emission performance in non-state-owned enterprises and high-tech industry enterprises compared to state-owned enterprises and non-high-tech industry enterprises. This may be due to the higher flexibility and autonomy of non-state-owned enterprises in rule-making, allowing for more flexible application of digital infrastructure and greater carbon reduction effects. On the other hand, high-tech industry enterprises enjoy more policy subsidies and technical support than non-high-tech industry enterprises, and they have advantages in talent and technology application. Therefore, the application of data elements such as digital infrastructure has a greater impact on high-tech industry enterprises, potentially promoting green product innovation and improving total factor carbon emission performance to a greater extent.

Based on the above conclusions, the study offers the following policy recommendations.

(1) Strengthen Digital Infrastructure Construction to Support Green Urban Transformation: Governments should promote the construction of next-generation digital infrastructure, such as 5G, big data, and artificial intelligence. By increasing investment in digital infrastructure, governments can stimulate domestic demand and drive economic growth. Expanding support for digital resources will amplify their role as substitutes for traditional inputs. Digital infrastructure significantly boosts urban total factor carbon emission performance. Given China’s industrial structure, economic growth tends to impact the environment; therefore, leveraging digital elements’ environmental advantages can help achieve economic growth while reducing pollution, aligning ecological and economic goals. For developing countries, enhancing digital infrastructure can foster both digital economic growth and green development within resource constraints. China is now combining digitization with computing power and electricity, putting forward policies such as “east digitization and west calculation” to address regional differences. It means digital infrastructure construction will be conducive to urban development in the new era, and combine digital infrastructure construction with other production factors may have a more far-reaching impact on the green development of cities.