ssrn-5854389
Reducing Urban Energy Intensity through Artificial Intelligence: Exploring the
Mediating Roles of New Energy and Green Industry in Chinese Cities
Xinwei Zhanga*
a College of Economics, Sichuan Agricultural University, Chengdu, 611130, China
A B S T R A C T
Artificial intelligence is increasingly recognized as a key driver in promoting industrial
intelligence and transforming urban energy systems. This study investigates the impact of artificial
intelligence development on energy intensity using panel data from 280 Chinese cities from 2011
to 2023. Empirical analyses reveal that artificial intelligence development significantly reduces
urban energy intensity, and this result remains robust after a series of robustness tests and
addressing endogeneity issues. Further mechanism analysis reveals that artificial intelligence
primarily promotes energy intensity reduction through three emerging pathways, including
supporting new energy industry development, facilitating green industry upgrading, and
stimulating green technological innovation. Moreover, artificial intelligence’s effect on reducing
energy intensity is stronger in cities implementing low-carbon pilot policies, possessing strong
digital economy policy support, and having well-developed green infrastructure, which
underscores the importance of favorable policies and infrastructure in enhancing artificial
intelligence’s energy efficiency benefits. Overall, these findings highlight artificial intelligence’s
vital enabling role in driving green and low-carbon urban transformation. Our analysis suggests
that policymakers should further promote deep integration of artificial intelligence with urban
energy systems and industrial sectors, optimize supportive policies, accelerate green infrastructure
construction, and advance intelligent development in new energy and green industries to enhance
energy efficiency and achieve sustainable urban development.
A R T I C L E I N F O
Keywords:
Artificial Intelligence
Energy Intensity
New Energy Industry
Green Industry
1. Introduction
Serving as an indispensable pillar for economic and social progress, activities of energy
production and consumption do not merely constitute the driving force behind modernization and
urbanization advancement but also stand as the central focus of contradictions amid global climate
change and environmental constraints. Energy consumption is a key contributor to carbon dioxide
(CO₂) emissions. The International Energy Agency (IEA) reports that global energy-related CO₂
emissions rose 0.8% in 2024 to a record 37.8 gigatonnes (Gt). Meanwhile, rapid economic and
population growth in emerging and developing economies is spurring surging energy demand and
related emissions. In 2024, these economies accounted for over 80% of global energy demand
growth, with their energy-related CO₂ emissions rising 1.5% to 375 million tonnes (Mt).
IEA statistics show China's 2024 energy demand rose approximately 3%, leading global
* Corresponding author. College of Economics, Sichuan Agricultural University, Chengdu,
611130, China ; https://orcid.org/0009-0005-7695-5801
E-mail address: zxw@sicau.edu.cn (X. Zhang).
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5854389
Preprint not peer reviewed
-- 1 of 19 --
absolute growth, while its energy-related carbon emissions comprised approximately 33.3% of the
global total. Extreme high temperatures drove China's energy-related CO₂ emissions up by
approximately 100 Mt that year (IEA, 2025). Consequently, China and the international
community face dual pressures from rising energy demand and green low-carbon transition.
Balancing the rising demand for economic growth with reductions in energy intensity (EI) and
advancing the green transition is a pressing priority for both combating climate change and
achieving the United Nations Sustainable Development Goals, particularly SDG7 (affordable and
clean energy) and SDG13 (climate action).
Responding to climate change and resource constraints, China launched its "dual-carbon"
strategy in 2020 to achieve carbon peak and neutrality. Key measures include reducing EI per unit
output and developing clean, low-carbon, secure, and efficient energy systems. However, energy
consumption and economic growth demonstrate strong interdependence; for example, electricity
use in urban industries acts as a key determinant and "barometer" of economic growth (Lu et al.,
2024). As the world's largest energy producer and consumer, China must balance economic
growth with conservation and emission reductions, yet its traditional extensive growth model and
consumption patterns remain major barriers.
China has positioned scientific and technological innovation as the core driver of its energy
transition, emphasizing energy-technology collaboration and advancing new-quality energy
production capabilities. This aims to build a safe, efficient energy system and foster a
consumption model of conservation, efficiency, sustainability and inclusiveness. In 2025, the
government released China's Energy Transition white paper, outlining two key measures that
leverage digital technologies for energy industry transformation and accelerate smart
infrastructure upgrades. Also in 2025, the "AI + Energy" High-Quality Development Guidelines
designated artificial intelligence (AI) innovation as the primary pathway for high-quality energy
development, targeting internationally leading energy-specific AI technologies and applications by
2030.
However, emerging research reveals AI's multifaceted impacts on energy systems. AI
optimizes production, enhances decision-making efficiency (Li et al., 2025c), promotes
technological diffusion and innovation (Liu et al., 2025; Zhong et al., 2025), facilitates intelligent
management (Zhang & Zeng, 2024), and demonstrates substantial EI reduction potential (Liu et
al., 2025). Conversely, AI implementation drives increased model training scales, surging
computing power demands (Vinuesa et al., 2020), and continuous data center expansion (De
Vries, 2023), collectively raising net energy consumption, with GPT-3's single training run
emitting CO₂ equivalent to 49 cars' annual emissions (Taddeo et al., 2021).
Meanwhile, AI-driven ICT and digital sector proliferation may increase energy demand
beyond efficiency gains from intelligent industrial transformation, elevating net consumption (Lin
and Zhou, 2025; Lange et al., 2020). Current studies on AI's EI impact remain largely confined to
enterprise or single-industry scales (Li et al., 2025c; Zhong et al., 2025). Rigorous research is
needed to elucidate AI's influence pathways for accurately gauging AI's effects on energy systems
and facilitating the sector's green, low-carbon transition. This study makes marginal contributions
in the following aspects:
First, existing research emphasizes enterprise or single-industry perspectives, lacking
systematic urban-level analysis of AI's energy-saving effects. Moreover, EI measurements in
literature depend on macro-level statistics with limited granularity. We employ panel data from
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5854389
Preprint not peer reviewed
-- 2 of 19 --
280 Chinese cities (2011–2023), combining nighttime satellite light and macro energy statistics
to systematically estimate city-level EI and examine AI development's effects. This enhances
scientific validity and urban-level resolution of EI measurement while creating a robust data
foundation for future studies on emerging technologies and city-level energy systems.
Second, this research examines differential effects of policies and infrastructure on AI's
energy-saving outcomes. Current literature explores heterogeneity through industrial or
geographic lenses (Liu et al., 2025; Zhong et al., 2025). In contrast, this paper employs subgroup
regressions using low-carbon city pilot status, digital economy policy attention, and green
infrastructure development to assess their effects on AI's energy-saving heterogeneity, uncovering
the critical role of policy environment and infrastructure in driving AI's green spillovers and
supporting digital-green policy pilots.
Third, this research uncovers novel mechanisms through which AI affects EI. Previous
studies examine direct effects or factors like management optimization and innovation capability
(Zhang & Zeng, 2024; Liu et al., 2021). This study investigates emerging pathways across new
energy sector development, green industry expansion, and green innovation promotion, examining
how AI empowers these domains, facilitates green patent generation, and drives sustained
reductions in EI, thereby extending the theoretical framework of AI-empowered energy
transformation and offering insights into AI's role in facilitating the green transition of urban
energy systems.
2. Literature review and theoretical analysis
2.1. Literature review
EI, defined as energy used per output (Vu & Demena, 2025), quantitatively measures
integrated energy efficiency. It indicates an economy’s energy dependence (Zhang et al., 2020)
and link to carbon emissions (Dunyo et al., 2024), playing a crucial role in balancing economic
growth with energy and environmental constraints. EI determinants are complex and layered.
Macro-level studies highlight the roles of low-carbon transition, globalization (Zaidi et al., 2024),
green technology transfer (Zhang et al., 2025a), and sustainable energy use (Feng et al., 2023). In
China, EI shifts are primarily driven by technological improvements (Huang et al., 2025),
industrial structure optimization, and a cleaner energy mix with greater non-fossil energy shares
(Dong et al., 2025). Micro-level EI changes arise from corporate governance, technological
innovation, environmental pressures, and policy, with key influences including managers’
experience (Sun et al., 2024), new technology adoption (Lin and Xu, 2024; Wang et al., 2025),
climate risk (Li et al., 2025a), green credit (Tan et al., 2025), and regulations (Chen et al., 2024;
Liu et al., 2023). Overall, technological and emerging technology innovations are pivotal in
lowering EI at both macro and micro levels (Zhang et al., 2025a; Vu & Demena, 2025).
Digital technologies have transformed economic, social, and energy systems, with their EI
impact gaining scholarly attention (Xiao et al., 2025; Hao et al., 2025). Digital innovations like
enterprise digital transformation and industrial robot deployment optimize production and reshape
energy use patterns, reducing EI (Guan et al., 2025; Matthess et al., 2023). Macro-level
mechanisms reveal digital technologies expand digital economy sectors (e.g., software and IT
services) and upgrade traditional industries, reducing high-energy industries' proportion and
energy consumption (Huang et al., 2023; Ma & Lin, 2025). Micro-level mechanisms reveal digital
applications reduce corporate EI by boosting productivity and innovation efficiency (Guan et al.,
2025), enabling energy-saving financing (Xiao et al., 2025), enhancing renewable innovation
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5854389
Preprint not peer reviewed
-- 3 of 19 --
awareness (Wu et al., 2025), scaling energy use economies (Hao et al., 2025), and optimizing
energy management (Hao et al., 2025).
Nevertheless, digital technologies' EI impact demonstrates uncertainty and complexity.
Superficial digitalization offers limited energy-saving potential, as generic solutions often yield
only short-term optimizations without iterative, enterprise-specific capability enhancements.
Moreover, the "rebound effect" (where efficiency gains stimulate heightened energy demand) may
reduce energy-saving outcomes from digital technologies (Guan et al., 2025). Consequently,
energy consumption growth from digital business expansion (e.g., digital equipment investment,
ICT hardware production) may exceed "energy-saving effects," increasing net consumption
(Lange et al., 2020; Matthess et al., 2023).
Recent AI-driven technological transformation has created novel pathways for industrial
digitalization and new opportunities for energy conservation and emissions reduction in
socio-economic systems (Fu et al., 2024). Collectively, AI-enabled industrial innovation and
intelligent transformation effectively reduce energy consumption and EI in corporate and
energy-intensive sectors (Zhang & Zeng, 2024; Liu et al., 2021). Macro-level transformation shifts
energy structures toward cleaner sources, improving economic system energy performance (Guo
et al., 2025). AI achieves energy reduction via four mechanisms, including enhanced enterprise
innovation efficiency (Liu et al., 2021), advanced intelligent management systems cutting costs
(Zhang & Zeng, 2024), substituted traditional production factors with optimized processes (Li et
al., 2025), and guided green technological upgrades (Liu et al., 2025). From the perspective of its
mechanisms, AI-driven energy reduction operates by enhancing operational innovation systems
(Liu et al., 2021), deploying intelligent management to reduce expenditures (Zhang & Zeng,
2024), substituting production factors with process optimization (Li et al., 2025), and advancing
green tech upgrading (Liu et al., 2025), collectively contributing to energy conservation.
Concurrently, AI facilitates technology diffusion and innovation efficiency gains (Zhong et
al., 2025), enhances regional human capital and green innovation capacity (Zhu et al., 2025), and
promotes renewable energy transitions (Wang, 2026), reducing energy consumption. However,
uncertainty persists regarding AI's EI impact due to energy-intensive model training (Vinuesa et
al., 2020), heightened computing demands from generative AI (Taddeo et al., 2021; Lin & Zhou,
2025), and accelerated digital service consumption through large-scale deployment (De Vries,
2023), constraining energy-saving potential. While research focuses on AI's conservation
potential, most evidence originates from enterprise or single-industry analyses (Liu et al., 2021; Li
et al., 2025; Zhong et al., 2025). Significantly, macro-scale investigations of AI's influence on EI,
particularly at urban levels, remain scarce, necessitating systematic macro-level evaluation to
verify reduction efficacy and advance mechanism exploration.
2.2. Theoretical analysis and research hypotheses
2.2.1 Direct Effects: The Role of AI in Lowering EI
As an emerging general-purpose technology, AI demonstrates broad applicability across
multiple industrial sectors while driving technological advancements and complementary
innovations within adopting industries (Teece, 2018). Given its broad integration across
socioeconomic systems, AI has attracted significant attention from multiple stakeholder groups,
specifically governments, industries, academia, and research institutions, as a pivotal catalyst for
global energy transition (Jin et al., 2020; Lyu and Liu, 2021).
Empirical evidence confirms AI's effectiveness in reducing EI and supporting energy system
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5854389
Preprint not peer reviewed
-- 4 of 19 --
decarbonization. AI improves energy efficiency and lowers EI by optimizing industrial and
enterprise production processes. Industrial robots and intelligent control systems adjust equipment
operation to match energy demands, minimizing idle or overload states and directly reducing
energy use per unit output (Liu et al., 2021; Li et al., 2025c). Additionally, AI reduces
human-induced downtime and errors, improves production continuity, and curbs intermittent
energy waste in production (Zhang & Zeng, 2024; Liu et al., 2025). AI also enables real-time
energy management optimization, further cutting EI. For instance, it optimizes renewable power
systems through intelligent forecasting and dynamic grid adjustment to raise efficiency (Liu et al.,
2022). Simultaneously, AI drives technological innovation and digital transformation, such as
energy-saving equipment R&D and IoT monitoring, helping upgrade technology and management
in energy systems, thus improving efficiency and reducing EI (Liu et al., 2025; Li et al., 2025c).
Although scholars recognize that AI may increase energy use through digital
technology-driven economic growth (Lange et al., 2020), this risk can be mitigated by
technological innovation and policy measures (Umar et al., 2025). In summary, AI holds
significant promise as an effective tool for reducing EI and consistently achieving lower EI across
broad applications.
H1: AI can effectively reduce EI.
2.2.2 AI Empowerment and EI Reduction: The Role of the New Energy Industry
The United Nations Conference on New Energy and Renewable Energy defines new energy
as technologically or developmentally innovative energy sources that support the global transition
from fossil fuels to a more diversified and sustainable energy mix (United Nations, 1981). Main
types include solar, geothermal, ocean, modern wind, biomass energy, and small hydropower. By
supporting the move to a green and sustainable structure, new energy development greatly reduces
EI. New energy primarily cuts EI by replacing high-consumption fossil fuels (Jiao et al., 2024).
For instance, Rahman et al. (2022) found that a unit increase in renewable energy share lowers
carbon intensity by 0.003 in 25 emerging economies and improves energy system efficiency.
Additionally, promoting new energy technologies is closely tied to transforming the energy
system’s structure. Raising the share of modern renewable energy in final energy use and
balancing global supply and demand (Xu and Lieuea, 2025) systematically lowers EI.
Furthermore, policy assessments show that encouraging new energy application effectively
reduces firms’ EI (Liu et al., 2023). Thus, new energy development is a key route to reducing EI.
As a transformative technology, AI accelerates new energy development and adoption. First,
AI enhances renewable energy system efficiency and application scale by forecasting photovoltaic
output and optimizing wind farm layouts, thus reducing unit costs (Alhasnawi et al., 2024;
Bennagi et al., 2024). Second, in the long run, AI increases the renewable share in the energy mix
by optimizing supply-demand balance and lessening reliance on traditional sources (Zhao et al.,
2024). AI also advances new energy technology innovation and speeds up technology iteration in
energy systems, promoting innovative applications (Zhang et al., 2024; Lee et al., 2025). Overall,
AI supports new energy advancement and helps lower EI.
H2: AI can reduce EI by facilitating the expansion of the new energy industry.
2.2.3 AI Empowerment and Reduction of EI: The Role of Green Industry
The United Nations Industrial Development Organization (UNIDO) defines Green Industry
as an industrial system focused on clean technologies, efficient resource use, low pollution, and
reduced energy consumption to promote both economic growth and environmental protection
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5854389
Preprint not peer reviewed
-- 5 of 19 --
(UNIDO, 2011).
Research shows that green industry development improves resource allocation, diversifies
and decarbonizes energy consumption, and advances the circular economy (Pegels and
Lütkenhorst, 2014; Ramalingam et al., 2025; Xie and Lin, 2025), supporting green energy
transition and EI reduction. AI, as an enabling technology, plays a key role in green industry
growth. In green manufacturing, AI allows for quantitative assessment of material and energy
efficiency as well as emissions (Mao et al., 2019), and enhances corporate ESG performance to
accelerate green transformation (Zhou et al., 2025). The Green Industrial Internet of Things
(GIIoT), central to industrial digitalization, integrates AI with renewables to build smart grid
ecosystems, improving efficiency (Tabaa et al., 2020). In green logistics, combining AI with
knowledge graphs enables real-time route and energy optimization using transportation and
environmental data (Samadhiya et al., 2025). In green supply chains, AI and IoT jointly optimize
inventory and transportation, reducing lifecycle energy use and emissions (Zhang et al., 2025). In
summary, AI development promotes green industry advancement, driving EI reduction.
H3: AI can reduce EI by promoting the development of green industry.
2.2.4 AI Empowerment and EI Reduction: The Role of Green Innovation
As a key force for sustainable development, green innovation extends beyond technological
and corporate transformation to underpin macro-level green, low-carbon, and high-quality growth.
First, green innovation directly improves micro-level production processes, boosting energy
efficiency through process and model upgrades for both rapid and sustained EI reductions (Li et
al., 2023; Wang et al., 2025). Second, it raises macro-level energy efficiency by optimizing
resource allocation, reducing dependence on fossil fuels, and promoting clean energy substitution
to increase overall system efficiency (Shahzad, 2025; Ullah et al., 2025).
Simultaneously, AI significantly promotes green innovation. At the micro level, AI improves
innovation resource allocation by easing financing constraints, accumulating human capital,
enhancing human-AI collaboration, and supporting data-driven decisions, thus optimizing
resource use and reducing information asymmetry to improve green innovation quality and
efficiency (Zhao et al., 2025; Chen et al., 2025; Li et al., 2025b; Guo et al., 2025). At the macro
level, AI acts as a digital connector, strengthening the synergy between green technology
innovation and renewable energy transition (Behera et al., 2025). It also fosters cross-domain
collaboration, increases R&D investment, and creates synergy effects in industrial chains,
improving collaborative efficiency and green innovation outcomes (Li et al., 2025; Zhao et al.,
2025; Chen et al., 2025). Consequently, AI development promotes green innovation, reducing EI.
H4: AI is capable of reducing EI through fostering green innovation.
Figure 1 illustrates a theoretical framework for AI's influence on EI.
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5854389
Preprint not peer reviewed
-- 6 of 19 --
Fig. 1. Theoretical framework.
3. Method and model
3.1. Data
This study examines AI’s impact on EI in Chinese cities. Based on data availability and
sample representativeness, we constructed a balanced panel dataset of 280 cities from 2011 to
2023 for regression analysis. AI data were obtained from the China Big Data Platform for
Social-Science (CBDPS). City-level energy consumption was estimated by combining nighttime
light data with provincial energy consumption, following Wu et al. (2014) and Chen et al. (2020a).
Provincial energy data were sourced from the China Energy Statistical Yearbook as well as
provincial and municipal statistical yearbooks. Nighttime light data were sourced from Harvard
Dataverse (Wu et al., 2021). We also collected socioeconomic data from the China City Statistical
Yearbook, CNRDS, and the China Public Policy and Green Development Database (CPPGD).
3.2. Variable
3.2.1. Dependent variable
As the primary metric, EI measures overall energy efficiency as energy use per unit of GDP,
reflecting how effectively energy inputs generate economic output. Unlike total energy use, EI
controls for economic scale, allowing for scientific comparison of energy efficiency across regions
and over time. Lower EI in cities indicates greater energy efficiency and progress toward intensive
green development.
EI was measured as the ratio of city-level energy consumption to real GDP. City-level energy
consumption was estimated by integrating provincial energy consumption data with nighttime
light data. Specifically, following Wu et al. (2014) and Chen et al. (2020), we established a linear
model between provincial total energy consumption and total nighttime light brightness to derive
annual light-energy conversion coefficients kit for each province i and year t, representing energy
consumption per unit light brightness. Subsequently, city-level energy consumption was derived
by applying the conversion coefficients kit to each city's nighttime light intensity during 2011 to
2023. Fig. 2(a,b) shows the spatiotemporal distribution of EI levels. During 2011-2023, most cities
declined in EI levels. In 2011, northeastern, northern, and northwestern regions exhibited higher
Literature
Review
And
Theoretical
Analysis
Artificial
Intelligence Energy Intensity Direct Effect
Energy Management
System Optimization
Artificial
Intelligence Energy Intensity Mechanism
Analysis
New Energy Industry
Green Industry
Green Innovation
Intelligent Optimization
of Production Processes
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5854389
Preprint not peer reviewed
-- 7 of 19 --
EI than central and southeastern coastal regions. By 2023, high EI areas decreased substantially,
with only northern cities maintaining high levels, while central and southeastern coastal regions
consistently registered low levels.
3.2.2. Core independent variable
Enterprises act as key drivers of emerging industries and are main implementers of new
technology. This study uses the number of AI enterprises as an indicator for regional AI
development. Fig. 2(c,d) displays the spatiotemporal patterns of AI levels. Between 2011 and
2023, most cities experienced notable growth in AI. In 2023, southeastern coastal, North China,
and East China regions showed higher AI development than northwest, northeast, and southwest
regions.
Fig. 2. Spatial distribution of EI and AI in 2011 and 2023.
3.2.3. Control variables
To rigorously assess AI’s impact on urban EI, this study includes control variables to address
confounders, with definitions and descriptive statistics shown in Table 1. A region’s economic
development level is likely to influence its technological innovation and energy use efficiency
(Guo et al., 2025; Umar et al., 2025), and is represented by real per capita GDP. Industrial
structure changes may affect EI by altering the share of high-energy sectors and resource
allocation efficiency (Luan et al., 2021), thus the proportion of tertiary industry in GDP quantifies
industrial structure. Foreign investment is likely to improve energy efficiency through technology
transfer and managerial expertise (De et al., 2021), measured by the number of foreign-invested
enterprises. Population size, assessed by the registered population, is likely to shape regional
energy consumption patterns (Huang et al., 2025). Government intervention may impact
technological innovation, infrastructure, and energy efficiency (Zhang et al., 2025c), measured by
local public budget expenditure. Financial support may affect the adoption of clean technologies
and resource allocation, thus influencing EI (Zhang et al., 2025b), with the year-end loan balance
of financial institutions used as its indicator.
Table 1
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5854389
Preprint not peer reviewed
-- 8 of 19 --
Definition of control variables
Statistical Summary Control
Variable Symbol Definition Mean Std. Min Max N
Economic
development PGDP Real Per Capita Regional Gross
Domestic Product 5.16 3.01 0.69 24.54 3640
Industrial
structure TSR Proportion of Tertiary Industry
Added Value to GDP 43.43 10.10 10.15 84.85 3640
Foreign
investment FIE Number of Foreign-Invested
Enterprises 94.06 285.51 0 3292 3640
Population size POP Registered Population 460.09 324.11 29.97 3416 3640
Government
intervention GEXP Local General Public Budget
Expenditure 480.07 716.96 56.69 9638.50 3640
Financial support FIN Year-end Balance of Loans from
Financial Institutions 4290.70 9123.00 91.20 108600 3640
3.2.4. Intermediary variables
This study selects the developmental levels of the new energy industry, green industry, and
green innovation as mediating variables to explore how AI affects EI, based on its theoretical
framework. The new energy industry’s development is assessed by the number of relevant
enterprises, with data from CBDPS. The green industry’s development is measured by firm counts
in sectors such as energy-saving equipment, advanced transportation manufacturing,
carbon-reduction retrofitting, and green low-carbon industrial transitions, using CPPGD data.
Green innovation is indicated by the number of granted green patents awarded per city, based on
CNRDS data. Definitions and descriptive statistics for these mediators are provided in Table 2.
Table 2
Definition of intermediary variables
Statistical Summary Intermediary
variables Symbol Definition Mean Std. Min Max N
New energy industry
development NEI The number of enterprises in
the new energy industry 1438.21 2809.68 11 38335 3640
Green industry
development GRI The number of enterprises in
the green industry 950.58 2089.26 5 23794 3640
Green innovation GPA The total number of
authorized green patents 483.80 1360.94 1 19373 3640
3.3. Econometric model
Based on the above variables and data, this study employs a panel data regression model to
investigate the impact of AI on EI and its underlying mechanisms, as specified below:
(1)
As specified in Equation (1), i denotes city and t denotes year. The dependent variable EIit
represents EI, defined as energy consumption per unit of GDP output. AIit, the core explanatory
variable, measures the level of AI development. Controlsit denotes a set of control variables, while
α0 to α2 are estimated parameters. The model includes individual fixed effects ui, time fixed effects
vt, and a stochastic error term εit. The coefficient α1 on AIit is the primary focus; it is hypothesized
to be negative, suggesting that AI development can reduce EI.
Given the limitations of the traditional three-step mediation test (Aguinis et al., 2017; Pieters,
0 1 2 it it it i t it EI AI Controls u v
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5854389
Preprint not peer reviewed
-- 9 of 19 --
2017), this study draws on the methodological framework of Chen et al. (2020b) and Li et al.
(2025c) to construct the mediation model:
(2)
Equation (2) specifies MVit as the mediator variable set, comprising the quantity of new
energy enterprises (NEI), green industry enterprises (GRI), and green patents (GPA). Parameters
β0 to β2 require estimation.
4. Empirical analysis
4.1. Analysis of typical facts
This study employs ridgeline plots to visualize kernel density distributions of EI and AI
enterprise stock, with results in Figure 3(a) and (b). Figure 3(a) shows EI distributions from
2011-2023. Leftward curve shifts indicate sustained EI decline and significant urban energy
efficiency gains in China. Increasing peaks and narrowing curves transition distributions from
"low-wide" to "high-narrow," implying cross-city convergence and balanced nationwide
reductions. Figure 3(b) presents AI enterprise stock distributions for 2011-2023. A log1p
transformation mitigates right-skewness while axis labels display original values. Rightward curve
shifts reflect accelerated AI enterprise growth. Peak morphology evolves from "high-narrow" to
"low-wide," showing expanded dispersion and widening regional disparities that suggest emerging
agglomeration effects. Collectively, decreasing EI and expanding AI enterprises exhibit coupled
spatiotemporal evolution, providing preliminary evidence that AI development may reduce EI and
supporting future empirical research.
Fig.3. Density Distribution of Energy Intensity and AI Enterprise Stock
4.2. Benchmark regression results
Table 3 reports baseline regression results examining AI's effect on EI. Column (1) shows a
pooled regression without controls, where lnAI coefficient is -0.1486 (significant at 1%). Column
(2) employs a two-way fixed effects model (city and year) without controls, yielding lnAI
coefficient of -0.0646 (significant at 1%), confirming AI significantly reduces EI after controlling
for temporal and regional heterogeneity. Column (3) adds controls for economic development
(lnPGDP), industrial structure (lnTSR), foreign investment (lnFIE), population (lnPOP),
government intervention (lnGEXP), and financial support (lnFIN) to the pooled model, where
lnAI coefficient remains -0.0658 and significant at 1%, affirming a stable AI impact on EI
reduction.
Table 3
0 1 2 it it it i t it MV AI Controls u v
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5854389
Preprint not peer reviewed
-- 10 of 19 --
The results of robustness regression
(1) (2) (3) (4)
lnAI -0.1486*** -0.0646*** -0.0658*** -0.0206**
(-32.6326) (-6.2664) (-7.4522) (-2.4774)
lnPGDP -0.5115*** -0.6449***
(-18.5063) (-27.8567)
lnTSR -0.1871*** 0.1387***
(-4.2161) (4.5477)
lnFIE -0.0169** -0.0110
(-2.2226) (-0.8809)
lnPOP -0.2999*** -1.0009***
(-12.9643) (-15.9877)
lnGEXP -0.0228 0.0441
(-0.8339) (1.5632)
lnFIN 0.1778*** 0.0006
(8.8219) (0.0302)
Constant 0.0711*** -0.1617*** 0.5098** 5.4482***
(4.8444) (-5.6537) (2.0631) (10.4199)
City FE NO Yes NO Yes
Year FE NO Yes NO Yes
Observations 3,640 3,640 3,640 3,640
R-squared 0.226 0.881 0.323 0.924
Note: Figures in parentheses are t-statistics. ***p < 0.01, **p < 0.05, *p < 0.1.
Column (4) results show a lnAI coefficient of -0.0206, significant at 5% level (p=0.013) and
approaching 1% significance. Among control variables, lnPGDP and lnPOP exhibit negative
coefficients significant at 1%, suggesting higher economic development and population size
correlate with lower EI, likely due to scale and agglomeration effects. Overall, by controlling for
variables and two-way fixed effects, Column (4) more precisely estimates the net AI effect.
Consequently, baseline regression results indicate AI development significantly lowers EI,
supporting Hypothesis 1.
4.3. Robustness checks
To ensure baseline reliability, robustness checks are summarized in Table 4. Column (1) uses
AI patents (lnAIP) as an alternative measure, with coefficient -0.0151 significant at 1%. Column
(2) applies 1% winsorization to variables, yielding lnAI coefficient of -0.0238 significant and
negative. Column (3) excludes 2020 data to eliminate COVID-19 impacts on EI, showing lnAI
coefficient -0.0182 significant at 5%. Column (4) employs province-year interaction fixed effects
to control for cross-provincial time trends, with lnAI coefficient -0.0415 significant at 1%.
Collectively, these checks demonstrate AI significantly reduces EI across variable substitution,
extreme value treatment, sample adjustment, and interactive fixed effects, confirming baseline
robustness.
Table 4
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5854389
Preprint not peer reviewed
-- 11 of 19 --
The results of robustness regression and endogenous amendment
(1) (2) (3) (4) (5) (6)
lnEI lnEI lnEI lnEI lnAI lnEI
lnAI -0.0238*** -0.0182** -0.0415*** -0.8554**
(-2.9052) (-2.1118) (-3.8973) (-2.3625)
lnAIP -0.0151***
(-5.2021)
IV 0.0710**
(2.4175)
Controls Yes Yes Yes Yes Yes Yes
City FE Yes Yes Yes NO Yes Yes
Year FE Yes Yes Yes NO Yes Yes
Province × Year FE Yes
Constant 5.1654*** 4.5690*** 5.4168*** 3.3264*** -3.0313*** /
(9.8597) (8.0778) (10.0576) (8.2654) (-2.6149) /
Observations 3,640 3,640 3,360 3,575 3,640 3640
R-squared 0.925 0.919 0.921 0.724 0.962 /
Note: Figures in parentheses are t-statistics. ***p < 0.01, **p < 0.05, *p < 0.1.
4.4. Endogenous Amendment
While baseline regressions show AI significantly reduces EI, endogeneity remains a concern.
Omitted variable bias may exist if unobserved factors influence both AI development and EI.
Reverse causality is also possible, as lower EI often results from industrial upgrading and
improved efficiency, which can promote high-tech industry agglomeration, including AI. To
address endogeneity, this study employs an instrumental variable approach and conducts
two-stage least squares (2SLS) estimation.
The instrumental variable (IV) is AI policy term frequency, measured by counting AI-related
policy references in city government work reports. On the one hand, AI mention frequency
reflects local policy support, implying subsidies and talent programs that promote AI
development, thus satisfying the IV relevance condition. First-stage results (Table 4, column 5)
show the IV significantly positively impacts AI. On the other hand, AI policy term frequency
reflects advocacy driven by strategy adjustments, not directly affecting EI. After controlling for
macroeconomic factors, no direct causal relationship exists between policy expressions and EI,
nor can it affect EI through other channels. Therefore, the IV meets the exogeneity requirement
and mitigates omitted variable bias and reverse causality. Column (6) in Table 4 shows the
second-stage lnAI coefficient is negative and significant at 5%, confirming AI's robust
EI-reducing effect after addressing endogeneity.
4.5. Heterogeneity analysis
To systematically analyze heterogeneity in AI's EI impact, this study performs subgroup
analyses across three dimensions (low-carbon pilot policies, digital economy policies, and green
infrastructure levels), with results detailed in Table 5.
4.5.1. Low-carbon cities policy
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5854389
Preprint not peer reviewed
-- 12 of 19 --
Based on China's low-carbon city pilot program list, sample cities are divided into pilot and
non-pilot groups. Table 5 columns (1)-(2) show that AI significantly reduces EI in pilot cities at
the 1% level, but has no significant effect in non-pilot cities. These findings suggest synergies
between low-carbon policies and AI enhance energy efficiency, likely due to pilot cities'
first-mover advantages in green transition frameworks and infrastructure, creating a supportive
environment for AI energy management applications.
4.5.2. Digital economy policy
Using CNRDS digital economy policy term frequency data, the sample is split into high- and
low-attention cities. Table 5, Columns (3)-(4) show the AI coefficient is -0.0396 and significant at
1% in high-attention cities, suggesting stronger digital economy policies amplify AI's green
spillover effects, reducing EI. However, it is insignificant in low-attention cities. These findings
highlight synergy between local digitalization strategies and AI's green empowerment. This likely
arises as digital economy policies accelerate AI integration with industries and energy systems,
enhancing the link between intelligent transformation and energy efficiency.
4.5.3. Green infrastructure
Using CPPGD green infrastructure enterprise data, this study performs subgroup regressions
for sample cities. The data cover building energy efficiency, green buildings, green transportation,
green logistics, and environmental infrastructure. Table 5 columns (5)-(6) show that in
high-green-infrastructure cities, AI's EI impact is -0.0303 and significant at 1%; the effect is
insignificant in low-green-infrastructure cities. This likely reflects that robust green infrastructure
enables efficient AI integration in energy systems, enhancing urban energy optimization
effectiveness.
Table 5
Tests of heterogeneity
Low-carbon cities policy Digital economy policy Green infrastructure
Yes No High Low High Low lnEI
(1) (2) (3) (4) (5) (6)
lnAI -0.0499*** 0.0056 -0.0396*** -0.0017 -0.0303*** -0.0111
(-3.9496) (0.5129) (-3.1442) (-0.1343) (-2.9509) (-0.9297)
Controls Yes Yes Yes Yes Yes Yes
City FE Yes Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes Yes
Constant 7.6871*** 3.8474*** 6.4935*** 3.2588*** 6.7644*** 2.0750**
(10.8312) (4.9725) (8.4791) (3.8230) (8.9026) (2.3719)
Observations 1,586 2,054 1,765 1,850 1,816 1,815
R-squared 0.937 0.912 0.939 0.922 0.933 0.913
Note: Figures in parentheses are t-statistics. ***p < 0.01, **p < 0.05, *p < 0.1.
4.6. Mechanism of action tests
This section analyzes AI's impact mechanisms on EI through new energy industry
development (NEI), green industry development (GRI), and green innovation (GPA). It uses
counts of new energy enterprises, green enterprises, and authorized green patents as mediators to
systematically evaluate AI's role in reducing EI.
4.6.1 New energy industry development
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5854389
Preprint not peer reviewed
-- 13 of 19 --
Table 6 column (1) shows that the lnAI coefficient for lnNEI is significantly positive at the
1% level, indicating AI significantly promotes new energy industry growth. AI improves new
energy system operations and renewable energy production efficiency, supporting the sector's
expansion (Bennagi et al., 2024; Alhasnawi et al., 2024). As the new energy industry expands, its
substitution for traditional energy strengthens, effectively limiting EI growth (Jiao et al., 2024;
Rahman et al., 2022). These results strongly support H2, which suggests AI reduces EI by
advancing new energy industry development.
4.6.2 Green industry development
Column (2) of Table 6 evaluates AI's effect on green industry development. For lnGRI, the
lnAI coefficient is 0.0584 and highly significant, indicating AI adoption fosters green industry
growth. AI improves resource and energy use, pollution control, and recycling in green
manufacturing, while enabling green logistics and supply chains (Mao et al., 2019; Zhou et al.,
2025). Such AI-driven green industry progress advances regional industrial upgrades and resource
optimization (Pegels & Lütkenhorst, 2014; Xie & Lin, 2025), thereby reducing EI. These results
strongly support H3.
4.6.3 Green innovation effect
As shown in Table 6, column (3), the AI regression coefficient for green patents is positive
and highly significant, indicating that AI significantly promotes green technological innovation.
By strengthening data analysis, R&D, and resource allocation, AI improves green innovation
efficiency at both firm and regional scales (Li et al., 2023; Guo et al., 2025). The growth in green
patents accelerates technology diffusion and application, helping firms save energy, lower
consumption, and increase renewable energy use, thereby reducing EI (Li et al., 2023; Wang et al.,
2025; Shahzad, 2025). These results support H4.
Table 6
Mechanism results.
(1) (2) (3)
lnNEI lnGRI lnGPA
lnAI 0.0696*** 0.0584*** 0.0598***
(6.2657) (7.1436) (2.7910)
Controls Yes Yes Yes
City FE Yes Yes Yes
Year FE Yes Yes Yes
Constant -0.2184 -1.2973*** 2.8504**
(-0.2922) (-2.7085) (2.2367)
Observations 3,640 3,640 3,640
R-squared 0.984 0.993 0.958
Note: Figures in parentheses are t-statistics. ***p < 0.01, **p < 0.05, *p < 0.1.
5.Conclusions and policy recommendations
5.1. Conclusions
Using panel data from 280 Chinese cities from 2011 to 2023, this study empirically examines
AI's impact on EI and its mechanisms through fixed-effects models, robustness tests, endogeneity
controls, and heterogeneity analyses. The main findings are as follows.
First, AI can significantly reduce urban EI, with results robust to checks. IV estimates
confirm this after addressing endogeneity, suggesting AI's practical value for energy efficiency
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5854389
Preprint not peer reviewed
-- 14 of 19 --
and green transformation. Second, AI's EI impact shows heterogeneity, with stronger effects in
low-carbon pilot cities, digital economy cities, and areas with green infrastructure. This suggests
favorable policy environments and infrastructure enhance AI's role in energy optimization. Third,
the mechanism analysis shows that AI can indirectly reduce EI through three emerging pathways.
AI promotes new energy development, advancing urban energy structures toward renewables; it
encourages green manufacturing and logistics, fostering greener urban industries with lower
energy use; and it drives green technological progress, improving urban energy efficiency.
5.2. Policy recommendations
Drawing on the research results, this study proposes four policy recommendations to guide
AI for EI reduction. First, enhance AI-energy system integration. Governments should improve
smart energy conditions and promote AI across all stages. Enterprises should increase AI R&D
using big data and algorithms to optimize renewables and smart grids, accelerating system
upgrades for green transformation and EI reduction. Second, strengthen policy-green
infrastructure coordination. Promote low-carbon cities and green infrastructure for AI green
empowerment, and improve green logistics, transportation, and smart grids to support AI-enabled
urban energy optimization and industry upgrading. Third, accelerate new energy and green
industry transformation. Encourage AI across new energy value chains and integrate AI with
green manufacturing and supply chains for low-carbon, intelligent upgrades. Finally, improve
green innovation support. Establish green technology innovation and IP protection mechanisms.
Encourage AI-driven green patent R&D and strengthen AI-enabled green innovation diffusion
through policy incentives.
CRediT authorship contribution statement
***: Writing–original draft, Data curation, Project administration, Visualization, Formal
analysis, Software, Writing–review & editing, Methodology, Conceptualization .
Declaration of competing interest
The authors declare that they have no known competing financial interests or personal
relationships that could have appeared to influence the work reported in this paper.
Acknowledgements
This research is supported by the China Postdoctoral Science Foundation (No.
2024M752299)
Data availability
Data will be made available on request.
References
Aguinis, H., Edwards, J.R., Bradley, K.J., 2017. Improving our understanding of moderation and mediation in
strategic management research. Organ. Res. Methods 20, 665–685.
Alhasnawi, B.N., Almutoki, S.M.M., Hussain, F.F.K., et al., 2024. A new methodology for reducing carbon
emissions using multi-renewable energy systems and artificial intelligence. Sustain. Cities Soc. 114, 105721.
Behera, B., Behera, P., Pata, U.K., et al., 2025. Artificial intelligence-driven green innovation for sustainable
development: Empirical insights from India's renewable energy transition. J. Environ. Manage. 389, 126285.
Bennagi, A., AlHousrya, O., Cotfas, D.T., et al., 2024. Comprehensive study of the artificial intelligence applied in
renewable energy. Energy Strategy Rev. 54, 101446.
Chen, J., Gao, M., Cheng, S., et al., 2020a. County-level CO2 emissions and sequestration in China during 1997–
2017. Sci. Data 7(1), 391.
Chen, L., Li, S., She, Z., 2025. A study on the impact of artificial intelligence applications on corporate green
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5854389
Preprint not peer reviewed
-- 15 of 19 --
technological innovation: A mechanism analysis from multiple perspectives. Int. Rev. Econ. Financ. 104490.
Chen, M., Xiao, H., Li, L., et al., 2024. How does government climate risk perception affect corporate energy
consumption and intensity? Energy Sustain. Dev. 81, 101496.
Chen, Y., Fan, Z., Gu, X., Zhou, L., 2020b. Arrival of young talent: The send-down movement and rural education
in China. Am. Econ. Rev. 110(11), 3393–3430.
De Vita, G., Li, C., Luo, Y., 2021. The inward FDI-energy intensity nexus in OECD countries: A sectoral R&D
threshold analysis. J. Environ. Manage. 287, 112290.
De Vries, A., 2023. The growing energy footprint of artificial intelligence. Joule 7(10), 2191–2194.
Dong, Y., Chen, T., Mao, J., et al., 2025. Greening through trading: The role of policy intervention on energy
intensity. Energy 314, 134193.
Dunyo, S.K., Odei, S.A., Chaiwet, W., 2024. Relationship between CO2 emissions, technological innovation, and
energy intensity: Moderating effects of economic and political uncertainty. J. Clean. Prod. 440, 140904.
Feng, Y., Zhang, J., Geng, Y., et al., 2023. Explaining and modeling the reduction effect of low-carbon energy
transition on energy intensity: Empirical evidence from global data. Energy 281, 128276.
Fu, Y., Shen, Y., Malin, S., et al., 2024. Does artificial intelligence reduce corporate energy consumption? New
evidence from China. Econ. Anal. Policy 83, 548–561.
Guan, R., Wang, H., Zheng, R., et al., 2025. Evaluating the impact of digital technologies on energy efficiency:
Evidence from Chinese publicly listed companies. Energy Policy 207, 114847.
Guo, L., Pei, H., Liu, Y., 2025. Artificial intelligence and corporate green innovation: Evidence from China. Res.
Int. Bus. Financ. 103039.
Hao, X., Miao, E., Sun, Q., et al., 2025. When climate policy's up in the air: How digital technology impacts
corporate energy intensity. Energy Econ. 144, 108311.
Huang, J., Wang, Y., Luan, B., et al., 2023. The energy intensity reduction effect of developing digital economy:
Theory and empirical evidence from China. Energy Econ. 128, 107193.
Huang, W., Miao, Y., Ye, H., et al., 2025. Trends and determinants of energy intensity in China: A study using
index decomposition and econometric analysis. Environ. Sustain. Indic. 100892.
International Energy Agency. (2025). Global Energy Review 2025.
https://www.iea.org/reports/global-energy-review-2025
Işık, C., Yan, J., Ongan, S., 2025. Energy intensity, supply chain digitization, technological progress bias in
China's industrial sectors. Energy Econ. 145, 108442.
Jiao, J., Song, J., Ding, T., 2024. The impact of synergistic development of renewable energy and digital economy
on energy intensity: Evidence from 33 countries. Energy 295, 130997.
Jin, D., Ocone, R., Jiao, K., et al., 2020. Energy and AI. Energy AI 1, 100002.
Lange, S., Pohl, J., Santarius, T., 2020. Digitalization and energy consumption. Does ICT reduce energy demand?
Ecol. Econ. 176, 106760.
Lee, C.C., Li, J., Yan, J., 2025. Can artificial intelligence contribute to the new energy system? Based on the
perspective of labor supply. Technol. Soc. 81, 102877.
Li, A., Ma, Y., Li, B., 2025a. How do climate risks affect corporate energy intensity? Evidence from China.
Energy 323, 135636.
Li, C., Zhang, Y., Liu, X., et al., 2025b. Does artificial intelligence promote green technology innovation in the
energy industry? Energy Econ. 144, 108402.
Li, H., Lu, Z., Zhang, Z., et al., 2025c. How does artificial intelligence affect manufacturing firms' energy
intensity? Energy Econ. 141, 108109.
Li, M., Yuan, N., Du, W., 2025d. Artificial intelligence and firm green innovation: Empirical evidence from the
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5854389
Preprint not peer reviewed
-- 16 of 19 --
application of robots in China. Econ. Anal. Policy 87, 2239–2253.
Li, Y., Cobbinah, J., Abban, O.J., et al., 2023. Does green manufacturing technology innovation decrease energy
intensity for sustainable development? Econ. Anal. Policy 78, 1010–1025.
Lin, B., Xu, C., 2024. The effects of industrial robots on firm energy intensity: From the perspective of
technological innovation and electrification. Technol. Forecast. Soc. Change 203, 123373.
Lin, B., Zhou, D., 2025. How does the explosive growth of AI affect China's power supply and demand: A
scenario simulation based on the LEAP model. Renew. Energy 124485.
Liu, L., Yang, K., Fujii, H., et al., 2021. Artificial intelligence and energy intensity in China’s industrial sector:
Effect and transmission channel. Econ. Anal. Policy 70, 276–293.
Liu, X., Cifuentes-Faura, J., Zhao, S., et al., 2025. Impact of artificial intelligence technology applications on
corporate energy consumption intensity. Gondwana Res. 138, 89–103.
Liu, X., Wang, C., Wu, H., et al., 2023. The impact of the new energy demonstration city construction on energy
consumption intensity: Exploring the sustainable potential of China's firms. Energy 283, 128716.
Lu, F., Ma, F., Hu, S., 2024. Does energy consumption play a key role? Re-evaluating the energy
consumption-economic growth nexus from GDP growth rates forecasting. Energy Econ. 129, 107268.
Luan, B., Zou, H., Chen, S., et al., 2021. The effect of industrial structure adjustment on China’s energy intensity:
Evidence from linear and nonlinear analysis. Energy 218, 119517.
Ma, R., Lin, B., 2025. The impact of digital technology innovation on energy-saving and emission reduction based
on the urban innovation environment. J. Environ. Manage. 375, 124176.
Mao, S., Wang, B., Tang, Y., et al., 2019. Opportunities and challenges of artificial intelligence for green
manufacturing in the process industry. Eng. 5(6), 995–1002.
Matthess, M., Kunkel, S., Dachrodt, M.F., et al., 2023. The impact of digitalization on energy intensity in
manufacturing sectors–A panel data analysis for Europe. J. Clean. Prod. 397, 136598.
Pegels, A., Lütkenhorst, W., 2014. Is Germany’s energy transition a case of successful green industrial policy?
Contrasting wind and solar PV. Energy Policy 74, 522–534.
Pieters, R., 2017. Meaningful mediation analysis: Plausible causal inference and informative communication. J.
Consum. Res. 44, 692–716.
Rahman, M.M., Sultana, N., Velayutham, E., 2022. Renewable energy, energy intensity and carbon reduction:
Experience of large emerging economies. Renew. Energy 184, 252–265.
Ramalingam, S., Hassan, W.H., Subramanian, M., et al., 2025. Synergies for sustainability: Renewable energy,
urban planning, and green industry in carbon emission reduction. Sustain. Futures 10, 101222.
Samadhiya, A., Kumar, A., Luthra, S., 2025. Integrating generative artificial intelligence into green logistics: A
systematic review and policy-oriented research agenda. J. Clean. Prod. 519, 146018.
Shahzad, U., 2025. Achieving clean energy transitions: How green innovation and financial development shape
energy usage. Energy Econ. 108851.
Sun, H., Han, F., Liu, Z., 2024. Does executive environmental protection experience reduce enterprise energy
consumption intensity? Econ. Anal. Policy 83, 652–666.
Tabaa, M., Monteiro, F., Bensag, H., et al., 2020. Green industrial internet of things from a smart industry
perspectives. Energy Rep. 6, 430–446.
Taddeo, M., Tsamados, A., Cowls, J., et al., 2021. Artificial intelligence and the climate emergency:
Opportunities, challenges, and recommendations. One Earth 4(6), 776–779.
Tan, W., Huang, Y., Chen, X.H., et al., 2025. Green credit policy and energy consumption intensity in China:
Firm-level evidence and insights. J. Environ. Manage. 373, 123664.
Teece, D.J., 2018. Profiting from innovation in the digital economy: Enabling technologies, standards, and
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5854389
Preprint not peer reviewed
-- 17 of 19 --
licensing models in the wireless world. Res. Policy 47(8), 1367–1387.
Ullah, U., Shaheen, W.A., Abdalkrim, G.M., et al., 2025. Past and present of energy: The role of green finance,
technological innovation, and financial risk in sustainability indicators. Environ. Sustain. Indic. 100936.
Umar, M., Horobet, A., Negreanu, C.C., et al., 2025. Fintech adoption, AI development, and energy transition
dynamics. Energy Econ. 108905.
United Nations Industrial Development Organization. (2011). Web Policies for Green Industry.
https://www.unido.org/sites/default/files/2011-05/web_policies_green_industry_0.pdf
United Nations. (1981). Report of the United Nations Conference on New and Renewable Sources of Energy,
Nairobi, 10 to 21 August 1981. https://digitallibrary.un.org/record/25034/?v=pdf
Vinuesa, R., Azizpour, H., Leite, I., et al., 2020. The role of artificial intelligence in achieving the Sustainable
Development Goals. Nat. Commun. 11(1), 233.
Vu, T.T.T., Demena, B.A., 2025. How does technology adoption affect energy intensity? Evidence from a
meta-analysis. Appl. Energy 398, 126439.
Wang, J., Duan, K., Zheng, Y., 2025. Green supply chain management, green technology innovation and firms'
energy consumption intensity. Energy Econ. 141, 108133.
Wang, Q., Zhang, S., Li, R., 2026. Artificial intelligence in the renewable energy transition: The critical role of
financial development. Renew. Sustain. Energy Rev. 226, 116280.
Wu, J.S., Niu, Y., Peng, J., et al., 2014. Research on energy consumption dynamic among prefecture-level cities in
China based on DMSP/OLS nighttime light. Geogr. Res. 33(4), 625–634.
Wu, L., Shi, J., 2025. From peer influence to green cognition: How digital transformation fosters renewable energy
innovation in manufacturing. Energy Econ. 108691.
Wu, Y., Shi, K., Chen, Z., et al., 2021. An improved time-series DMSP-OLS-like data (1992–2023) in China by
integrating DMSP-OLS and SNPP-VIIRS. Harvard Dataverse, V6. https://doi.org/10.7910/DVN/GIYGJU.
Xiao, Y., Duan, Y., Zhou, H., et al., 2025. Has digital technology innovation improved urban total factor energy
efficiency?—Evidence from 282 prefecture-level cities in China. J. Environ. Manage. 378, 124784.
Xie, J., Lin, B., 2025. Evaluating the role of green industrial policy in enhancing energy efficiency: A
quasi-natural experiment based on the Five-Year Plans in China. Process Saf. Environ. Prot. 107902.
Xu, X., Lieuea, B., 2025. Contribution of modern renewables to final energy consumption, energy intensity, and
environmental quality: The role of world energy balances, economic performance, and shadow economies.
Energy 137191.
Zaidi, S.A.H., Ashraf, R.U., Hassan, T., 2024. Effects of globalization and financial inclusion on energy intensity:
The case of emerging economies. Energy 306, 132380.
Zhang, B., Tian, X., He, B., et al., 2025a. The impact of China's green technology transfer on energy intensity in
countries along the Belt and Road. J. Environ. Manage. 373, 123691.
Zhang, C., Liu, J., Abedin, M.Z., et al., 2025b. Navigating sustainable finance: Examining the impact of
sustainable credit policy on energy consumption intensity. Res. Int. Bus. Financ. 73, 102594.
Zhang, C., Su, B., Zhou, K., et al., 2020. A multi-dimensional analysis on microeconomic factors of China's
industrial energy intensity (2000–2017). Energy Policy 147, 111836.
Zhang, H., Lv, Y., Zhang, J.Z., et al., 2025c. Government science and technology spending and energy efficiency:
An inverted U-shaped analysis. Energy Econ. 108818.
Zhang, L., Innab, N., Shuhidan, S.M., et al., 2025d. Artificial intelligence-driven internet of things-based green
supply chain for carbon reduction in sustainable manufacturing. J. Environ. Manage. 389, 126170.
Zhang, W., Zeng, M., 2024. Is artificial intelligence a curse or a blessing for enterprise energy intensity? Evidence
from China. Energy Econ. 134, 107561.
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5854389
Preprint not peer reviewed
-- 18 of 19 --
Zhang, X., Khan, K., Shao, X., et al., 2024. The rising role of artificial intelligence in renewable energy
development in China. Energy Econ. 132, 107489.
Zhao, M., Fu, X., Sun, J., et al., 2025. Optimal strategy of artificial intelligence on low-carbon energy
transformation: Perspective from enterprise green technology innovation efficiency. Energy 319, 135035.
Zhao, Q., Wang, L., Stan, S.E., et al., 2024. Can artificial intelligence help accelerate the transition to renewable
energy? Energy Econ. 134, 107584.
Zhong, C., Cai, H., Fang, S., et al., 2025. Does artificial intelligence reduce energy intensity in manufacturing?
Evidence from country-level data. Energy Econ. 108784.
Zhou, C., Zhang, H., Ying, J., et al., 2025. Artificial intelligence and green transformation of manufacturing
enterprises. Int. Rev. Financ. Anal. 1043
Zhu, Q., Che, J., Liu, S., et al., 2025. How can artificial intelligence technology applications accelerate energy
innovation in China? Evidence from provincial regional data. Econ. Anal. Policy.
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5854389
Preprint not peer reviewed
-- 19 of 19 --