Unpacking Artificial Intelligence’s Role in the Energy Transition: The Mediating and Moderating Roles of Knowledge Production and Financial Development

Abstract

This study pioneers an investigation into how artificial intelligence (AI)—shaped by financial development and knowledge production—is transforming the energy transition across BRICS economies and paving the way for a digitally enabled, sustainable future. Using panel data for 2005–2020, the findings confirm that AI is the primary driver of both explicit (EET) and implicit (IET) energy transitions in the BRICS nations, while economic growth, human capital, and financial globalization play comparatively smaller roles. We further find that AI’s effect on the explicit transition is fully mediated by efficiency gains. Financial development weakens—whereas knowledge production strengthens—AI’s green impact. Robustness checks across alternative models support these results, and spillover analyses indicate that cross-border AI advances, economic growth, human capital, and innovation flows shape each BRICS country’s energy-transition path. Based on these findings, the study proposes coordinated policy packages to harness AI for the energy transition while managing distributional and cross-border effects.

1. Introduction

The BRICS nations are on the frontline of climate change, experiencing accelerated warming, extreme weather, and mounting sea-level rise that threaten food security, water supplies, and urban infrastructure [1,2]. With coal and oil still underpinning the bulk of their energy mixes—Russia and South Africa remain deeply carbon-dependent “carbon states,” while China and Brazil are emerging as “electrostates” with growing clean-energy capacity—there is an urgent need to pivot toward low-carbon pathways to meet both climate goals and sustainable development targets (https://www.ft.com/content/5cbdcbd0-60ce-46ec-9582-f5ebf8d02825; accessed on 5 May 2025). According to the BRICS Policy Center’s Climate Ambition report (https://bricspolicycenter.org/wp-content/uploads/2024/11/Climate-Ambition-of-BRICS-countries-web-version.pdf; accessed on 24 June 2025), over 40% of the bloc’s collective energy supply is now renewable, yet policy fragmentation and continued fossil-fuel subsidies in some members undermine unified progress toward net-zero emissions Climate Ambition of BRICS Countries. Investment in the energy transition has soared globally, breaking the USD 3 trillion threshold for the first time in 2024, with USD 2 trillion directed toward clean-energy technologies and infrastructure (https://www.iea.org/reports/world-energy-investment-2024/overview-and-key-findings; accessed on 30 July 2025). Within the BRICS countries, China leads the pack—accounting for over 50% of the region’s renewables capital—followed by India’s rapid scale-up of utility-scale solar and wind projects; Brazil and South Africa have mobilized several billion dollars each in grid modernization and microgrid initiatives, while Russia lags but has recently announced a USD 10 billion green-hydrogen roadmap (https://globalenergymonitor.org/report/energy-in-the-brics/; accessed on 16 April 2025). These flows underpin explicit transitions (EET), such as increasing the share of renewables from 15% to 30% of total generation, as well as implicit transitions (IET), including investments in smart-grid automation and energy-efficiency retrofits that reduce transmission losses by up to 10%. However, the BRICS nations face persistent and multifaceted challenges—including insufficient and aging grid infrastructure, critical minerals bottlenecks that constrain clean technology manufacturing, chronic financing shortfalls in developing members, and policy incoherence stemming from divergent national priorities—that collectively threaten to stall, dilute, or even reverse recent gains. Without coordinated action, strategic investment alignment, and targeted reforms, these structural barriers could slow the momentum toward both explicit and implicit energy transitions, jeopardizing the bloc’s ability to meet its climate and sustainable development commitments.

The intersection between artificial intelligence (AI) and energy transition represents a transformative synergy with the potential to accelerate the shift toward cleaner and more sustainable energy systems. AI, through its capabilities in data analytics, optimization, and predictive modeling, plays a pivotal role in enhancing energy efficiency, managing decentralized energy systems, and supporting the integration of renewable energy sources into power grids. For instance, AI technologies such as machine learning algorithms are employed to forecast renewable energy generation from variable sources like wind and solar, thereby improving grid stability and reducing reliance on fossil fuels [3,4]. These technologies can also optimize the operation of smart grids, balance demand and supply in real time, and facilitate demand-side management through smart meters and intelligent control systems [5]. Thus, AI not only contributes to the operational efficiency of energy systems but also supports long-term planning and policy modeling for sustainable transitions. Moreover, AI contributes to innovation in clean energy technologies by enabling more efficient research and development processes. It accelerates the discovery of new materials for energy storage and photovoltaic systems by analyzing large datasets that would be too complex for traditional computational methods [6]. In the industrial sector, AI applications in predictive maintenance and process optimization lead to reduced energy consumption and emissions, thereby contributing to the broader goals of energy transition [7]. However, the deployment of AI must also be aligned with ethical considerations and energy justice principles, particularly regarding the energy consumption of data centers and the digital divide in AI access [8]. In sum, AI serves as both a driver and enabler of energy transition by improving the efficiency, resilience, and sustainability of energy systems.

Financial development significantly influences the relationship between artificial intelligence (AI) and energy transition by enhancing access to capital, reducing investment risks, and enabling the financing of AI-driven clean energy technologies. Well-developed financial systems facilitate the deployment of AI in areas such as smart grids, energy forecasting, and efficiency optimization, thereby accelerating both explicit transitions (towards renewables) and implicit transitions (via improved energy efficiency) [2]. Studies show that in regions like the EU and ASEAN, financial development strengthens the impact of AI on energy transition [9], while in emerging markets, it determines the scale and effectiveness of AI integration into energy systems. Conversely, in countries with underdeveloped financial sectors, capital constraints and policy uncertainty can limit AI adoption and stall energy transition efforts [10]. Thus, financial development acts as a vital enabler that conditions how effectively AI can drive sustainable energy transformation.

Moreover, knowledge production can significantly shape the relationship between artificial intelligence (AI) and energy transition by providing the scientific foundation and technical expertise necessary for AI-driven clean energy innovations. Through investments in research and development, academic output, and innovation systems, countries can design and implement AI applications—such as energy forecasting, smart grid optimization, and predictive maintenance—that support both explicit and implicit energy transitions. Studies show that high levels of knowledge production enhance the effectiveness of AI in reducing emissions and improving energy efficiency [11,12]. In contrast, limited knowledge production can hinder the adaptation and deployment of AI technologies, weakening their impact on sustainable energy systems [13]. Thus, knowledge production acts as a critical enabler, ensuring that AI tools are effectively developed and applied to drive meaningful progress in energy transition. Based on the above discussion, the study seeks to answer the research questions by formulating the following main research objectives:

(a)

To evaluate the direct impact of artificial intelligence on both explicit and implicit energy transitions in BRICS nations.

(b)

To investigate the cross-country spillover effects of artificial intelligence on domestic energy transitions.

(c)

To examine the mediating role of implicit energy transition in the relationship between AI and explicit energy transition.

(d)

To assess the moderating influence of financial development and knowledge production on the AI–energy transition nexus.

Novelty and Contribution of Study

This study makes three key contributions to the energy-transition literature. First, it extends [14] by rigorously decomposing the energy-transition process into its explicit (energy-mix shifts) and implicit (system-efficiency gains) components and empirically examining how artificial intelligence drives each dimension. Furthermore, we introduce a richer set of covariates—urbanization, financial globalization, human capital, technological innovations and economic risk—to control for broader macroeconomic and institutional influences, thereby addressing potential omitted-variable bias in prior work. Finally, by integrating these additional controls into the AI–energy-transition framework, we fill a critical gap in the literature and offer a more nuanced discourse on the conditions under which AI can most effectively catalyze low-carbon transformations across emerging economies.

Secondly, by focusing on the BRICS bloc, this study embeds financial development and knowledge production as both mediators and moderators of AI’s dual energy-transition pathways. In China and India—where green-bond markets and mandatory green-loan quotas are well established—deep financial systems direct AI-enabled capital into renewable-energy projects and efficiency upgrades, whereas in Brazil and South Africa, less developed green-finance infrastructures allow AI investments to reinforce existing fossil-fuel sectors. Similarly, robust R&D ecosystems and high patent output in China and India bolster AI’s capacity to generate both system-level efficiency gains and large-scale shifts in the energy mix, while weaker innovation environments in Russia, Brazil, and South Africa constrain AI’s transformative potential. This BRICS-specific framework demonstrates that identical AI can produce markedly different outcomes in explicit and implicit energy transitions, driven by each country’s financial structures and knowledge capacities.

Third, this study pioneers the analysis of AI’s cross-border spillovers within the BRICS countries, showing how peer-country AI intensity shapes both explicit renewable consumption and behind-the-meter efficiency gains (IET), all while rigorously controlling for human capital, financial globalization, technological innovation, and economic growth. Employing two-way fixed-effects regressions with average peer AI measures, we demonstrate that surges in AI investment abroad can draw away critical financial and human resources—undermining domestic energy-mix shifts and efficiency improvements—or, under certain conditions, spark demonstration effects that accelerate local transitions. By isolating these spillover dynamics from core macroeconomic drivers, our work forges a new discourse on transnational technology diffusion in energy systems and highlights the strategic importance of BRICS-wide coordination in AI deployment to maximize collective progress toward low-carbon futures.

The remainder of this paper is organized as follows: Section 2 reviews the theoretical framework and related literature; Section 3 describes the data and methodology; Section 4 presents the empirical results; and Section 5 offers conclusions and policy implications.

2. Theoretical and Literature Review

2.1. Theoretical Review

Artificial intelligence (AI) functions as a general-purpose technology that directly drives both explicit shifts toward renewable energy (EET) and behind-the-meter efficiency gains (IET) by optimizing grid operations, enabling predictive maintenance, and enhancing demand-response management [15]. According to the diffusion of innovation theory, these process innovations diffuse unevenly across countries, creating competitive dynamics wherein a surge in AI adoption by one BRICS member can attract scarce financial and human resources away from its peers—resulting in negative spillovers on both EET and IET in neighboring nations [16]. Mediation theory further suggests that AI’s most potent lever for boosting visible energy-mix transitions is its capacity to first improve underlying system efficiencies (IET), which then catalyze structural changes in the energy portfolio [3,14].

The strength and direction of AI’s impact on energy transitions are, however, contingent on each country’s financial and innovation infrastructure. Financial development (FD) can either reinforce or undermine AI’s green potential: under robust green-finance regimes, deeper capital markets lower the cost of AI-enabled clean-energy investments and amplify AI’s direct effects on EET and IET [9,17], whereas in the absence of targeted incentives, mature financial systems may crowd out renewables by favoring entrenched fossil-fuel incumbents [18,19]. Similarly, knowledge production (KP)—measured by patent intensity and R&D volume—serves as a nonlinear moderator: only after surpassing critical knowledge-stock thresholds do synergistic effects with AI intensify, accelerating both efficiency enhancements and large-scale renewable uptake [14]. Together, these moderating mechanisms illustrate that AI’s transformative power in BRICS energy transitions hinges on the interplay between technological adoption and the surrounding financial and knowledge ecosystems.

2.2. Literature Review

Over the past several years, research has consistently shown that advances in artificial intelligence (AI) play a key role in accelerating the energy transition (ET). The authors of Ref. [8] developed the AI-Enabled Energy Model (AIEM) and found that increased AI deployment corresponds with greater adoption of renewable energy and improved system efficiencies across various countries. Complementing this, the authors of Ref. [20] conducted a bibliometric analysis of 469 publications from 2006–2022, documenting the field’s growth and its growing orientation toward data-driven decision-making frameworks. Building on these insights, the authors of Ref. [4] applied machine-learning techniques to demonstrate that AI-based predictive and optimization tools directly facilitate integration of intermittent energy sources. Meanwhile, the authors of [3], using Baron and Kenny’s mediation approach with data from 15 emerging economies (2015–2022), showed that AI exerts both a direct effect on ET metrics and an indirect influence by bolstering institutional capacity and investment flows. At the empirical frontier, panel regressions and causal inference methods further substantiate the causal AI and ET relationship. The authors of [5] employed fixed-effects models for BRICS nations (2005–2019), finding that incremental AI investments lead to measurable improvements in renewable output and reductions in carbon intensity. In the United States, the authors of Ref. [21] used benchmark regression on 2018–2021 data to document a significant positive impact of AI deployment on clean energy generation and system-wide efficiency gains.

Over the past two decades, a substantial thread of research has portrayed financial development as a catalyst for the clean-energy transition by mobilizing capital and lowering financing costs for renewables. In their FMOLS study of 32 EU and ASEAN nations (2000–2020), the authors of Ref. [9] found that deeper credit markets and more liquid stock exchanges correlate with higher shares of renewable energy generation. Likewise, the authors of Ref. [22] applied a fixed-effects model to BRICS economies (1995–2022) and concluded that expanded banking sector depth and equity market activity significantly boost energy transition indices. The authors of Ref. [23], using panel quantile regression across developing markets (2000–2021), showed that the positive impact of financial development intensifies among countries already exhibiting strong renewable uptake, suggesting that well-capitalized finance sectors can amplify existing green momentum.

A contrasting literature strand warns that financial deepening may, under certain conditions, slow the shift away from fossil fuels. The authors of Ref. [10] employed fixed and random effects estimators on seven emerging economies (1990–2019) and document cases in which wider access to credit has instead funneled resources into carbon-intensive industries, thereby retarding renewable adoption. Applying panel ARDL techniques to 119 countries (1996–2019), the authors of [24] similarly revealed that without robust environmental regulation, enhanced financial intermediation can reinforce traditional energy pathways, yielding a net negative effect on transition metrics. The authors of Ref. [25] used system GMM for African nations (2006–2019) to uncover a non-linear relationship: financial development initially supports clean energy investment but eventually diverts funds back to familiar, high-carbon sectors.

Human capital and economic growth offer another dimension to the transition debate, with scholars probing whether better-educated workforces and rising incomes accelerate or complicate renewable energy deployment. The authors of Ref. [11] deployed system GMM on a panel of 134 countries (1996–2019) and detected an inverted-U relationship: at low income levels, gains in GDP per capita and human capital have limited impact on clean energy shares, but beyond a threshold—once basic infrastructure and skills are in place—their influence becomes markedly positive. This pattern suggests that growth unaccompanied by targeted education and training may fail to translate into greener energy systems. Complementary evidence, however, paints a more uniformly positive role for human capital and economic growth in driving energy transition. The authors of Ref. [13] used FMOLS on eight Asian nations (1995–2018) to show that improvements in literacy rates and GDP per capita robustly raise renewable energy penetration. The authors of Ref. [26] applied panel quantile regression to emerging economies (1990–2021), finding that the impact of human capital is especially strong at lower quantiles of energy transition performance, signaling an equity dimension to clean-energy diffusion. The authors of Ref. [27], via Panel Corrected Standard Errors for SAARC countries, and the authors of Ref. [28], using FMOLS on OECD nations (1980–2015), both confirm that better-educated populations and expanding economies consistently spur renewable adoption. Collectively, these studies argue that coupling growth strategies with investments in education creates a virtuous cycle for decarbonization.

Technological innovation and globalization emerge as powerful levers in unlocking the energy transition by reducing costs and facilitating technology diffusion. The authors of Ref. [29] applied CUP-FM and CUP-BC estimators to G7 countries (1985–2018) and demonstrated that R&D spending and trade openness mutually reinforce renewable capacity expansion. The authors of Ref. [30], using a bootstrap rolling-window approach for the UK (1995–2020), showed that periods of intensified global integration coincide with spikes in innovation-driven clean-energy outcomes. The authors of [31] employed DOLS on MENA economies (1997–2021), confirming that globalization magnifies the effect of domestic technological R&D on renewable generation. In a complementary SWOT analysis, the authors of Ref. [32] highlighted that international partnerships and cross-border R&D collaborations are essential for overcoming local capacity constraints and scaling up clean-tech deployment. Table 1 presents summary of reviewed studies.

Table 1. Synopsis of past studies.

Author(s)	Period	Nation(s)	Method(s)	Finding(s)
Artificial Intelligence and Energy Transition
[8]	Undefined	Global	AIEM	AI ↑ ET
[20]	2006–2022	469 publications	Bibliometric analysis	AI ↑ ET
[4]	Undefined	Undefined	Machine learning	AI → ET
[3]	2015–2022	15 emerging nations	Baron and Kenny’s approach	AI ↑ ET
[14]	2005–2019	BRICS nations	Two-way fixed effects regression	AI ↑ ET
[21]	2018–2021	U.S.	Benchmark regression	AI ↑ ET
Financial Development and Energy Transition
[9]	2000–2020	32 EU and ASEAN nations	FMOLS	FD ↑ ET
[22]	1995–2022	BRICS	Fixed Effect Model	FD ↑ ET
[23]	2000–2021	Developing markets	panel quantile regression	FD ↑ ET
[10]	1990–2019	7 emerging nations	Fixed Effects and Random Effects	FD ↓ ET
[24]	1996–2019	119 nations	panel ARDL	FD ↓ ET
[25]	2006–2019	Africa	GMM	FD ↑↓ ET
Human Capital, Economic Growth, and Energy Transition
[11]	1996–2019	134 countries	SGMM	HC and EG ↑↓ ET
[13]	1995–2018	8 Asian nations	FMOLS	HC and EG ↑ ET
[26]	1990–2021	Emerging nations	Panel Quantile Regression	HC and EG ↑ ET
[27]	Undefined	SAARC nations	PCSE	HC and EG ↑ ET
[28]	1980–2015	OECD nations	FMOLS	HC and EG ↑ ET
Technological Innovation, Globalization, and Energy Transition
[29]	1985–2018	G7 countries	CUP-FM and CUP-BC	TI and GLO ↑ ET
[30]	1995–2020	UK	bootstrap rolling window approach	TI and FIG ↑ ET
[31]	1997–2021	MENA economies	DOLS	TI and GLO ↑ ET
[32]	Not defined	Global analysis	SWOT analysis	TI and GLO ↑ ET

Note: ↑ and ↓ denote increase and decrease, respectively, and → represents one-way causality.

Table 1. Synopsis of past studies.

Author(s)	Period	Nation(s)	Method(s)	Finding(s)
Artificial Intelligence and Energy Transition
[8]	Undefined	Global	AIEM	AI ↑ ET
[20]	2006–2022	469 publications	Bibliometric analysis	AI ↑ ET
[4]	Undefined	Undefined	Machine learning	AI → ET
[3]	2015–2022	15 emerging nations	Baron and Kenny’s approach	AI ↑ ET
[14]	2005–2019	BRICS nations	Two-way fixed effects regression	AI ↑ ET
[21]	2018–2021	U.S.	Benchmark regression	AI ↑ ET
Financial Development and Energy Transition
[9]	2000–2020	32 EU and ASEAN nations	FMOLS	FD ↑ ET
[22]	1995–2022	BRICS	Fixed Effect Model	FD ↑ ET
[23]	2000–2021	Developing markets	panel quantile regression	FD ↑ ET
[10]	1990–2019	7 emerging nations	Fixed Effects and Random Effects	FD ↓ ET
[24]	1996–2019	119 nations	panel ARDL	FD ↓ ET
[25]	2006–2019	Africa	GMM	FD ↑↓ ET
Human Capital, Economic Growth, and Energy Transition
[11]	1996–2019	134 countries	SGMM	HC and EG ↑↓ ET
[13]	1995–2018	8 Asian nations	FMOLS	HC and EG ↑ ET
[26]	1990–2021	Emerging nations	Panel Quantile Regression	HC and EG ↑ ET
[27]	Undefined	SAARC nations	PCSE	HC and EG ↑ ET
[28]	1980–2015	OECD nations	FMOLS	HC and EG ↑ ET
Technological Innovation, Globalization, and Energy Transition
[29]	1985–2018	G7 countries	CUP-FM and CUP-BC	TI and GLO ↑ ET
[30]	1995–2020	UK	bootstrap rolling window approach	TI and FIG ↑ ET
[31]	1997–2021	MENA economies	DOLS	TI and GLO ↑ ET
[32]	Not defined	Global analysis	SWOT analysis	TI and GLO ↑ ET

Note: ↑ and ↓ denote increase and decrease, respectively, and → represents one-way causality.

2.3. Hypothesis Formulation

Harnessing the power of artificial intelligence, our study presents a bold roadmap for how “smart minds” reshape BRICS energy futures. Below are the nine targeted hypotheses (see Figure 1) that unpack AI’s direct, spillover, mediating, and moderating roles in driving both visible and efficiency-based energy transitions.

Hypothesis 1:

(Direct Effect on EET): Artificial intelligence has a significant effect on explicit energy transition.

Hypothesis 2:

(Direct Effect on IET): Artificial intelligence has a positive and significant effect on implicit energy transition.

Hypothesis 3:

(Spillover on EET): Peer-country AI exerts a significant spillover effect on domestic explicit energy transition.

Hypothesis 4:

(Spillover on IET): Peer-country AI intensity exerts a significant (negative) spillover effect on domestic implicit energy transition.

Hypothesis 5:

(Mediation via IET): Implicit energy transition mediates the impact of AI on explicit energy transition.

Hypothesis 6:

(Moderation by Financial Development on EET): Financial development moderates the effect of AI on explicit energy transition.

Hypothesis 7:

(Moderation by Financial Development on IET): Financial development moderates the effect of AI on implicit energy transition.

Hypothesis 8:

(Moderation by Knowledge Production on EET): Knowledge production moderates the effect of AI on explicit energy transition.

Hypothesis 9:

(Moderation by Knowledge Production on IET): Knowledge production moderates the effect of AI on implicit energy transition.

3. Data and Method

3.1. Data

This study investigates the key drivers of both explicit (shifts in the energy mix toward renewables) and implicit (system-level efficiency gains) energy transitions across the BRICS nations over the 2005–2020 period. Our primary dependent variables are explicit energy transition (EET) and implicit energy transition (IET), while we explore the moderating influence of knowledge production (KP) and financial development (FD) on these relationships. To ensure a comprehensive analysis, we include a set of controls—human capital (HC), economic growth (EG), financial globalization (FIG), technological innovation intensity (TI), urbanization (UB), and economic risk (ER)—each of which is precisely defined and sourced in Table 2. Employing a fixed-effects panel framework augmented by mediation and interaction models, we rigorously disentangle how innovation ecosystems and financial market depth shape the pace and nature of energy transitions within emerging economies.

Table 2. Data and measurement.

Role	Variables	Dimension	Index Interpretation	Sources
Dependent Variables	Explicit Energy Transition (EET)	Energy–Environment	Ratio of non-fossil generation to total generation	[33]
		Energy–Environment-Society	Share of population with clean cooking fuels
		Energy–Environment–Technology	Share of GDP spent on R&D
		Energy–Environment–Technology	Patent-based measure of clean energy innovation
	Implicit Energy Transition (IET)	Energy–Environment–Society	Average energy consumption per capita	[33]
		Energy–Environment–Economy	Energy use per unit of GDP
		Energy–Environment–Politics	Electricity imports/consumption ratio
		Energy–Environment–Politics	Modern renewables’ share in final consumption
		Energy–Environment–Economy	CO2 emissions per unit of energy supply
Moderating Variables	Knowledge Production (KP)		Number of papers published, the number of papers cited, and the citation count	[33]
	Financial Development (FD)		Domestic credit to private sector by banks (% of GDP)	[34]
Control Variables	Human Capital (HC)		Index	[35]
	Economic Growth (EG)		GDP per capita (constant 2015 US$)	[34]
	Financial Globalization (FIG)		Index	[36]
	Technological Innovations (TI)		Patents Resident	[33]
	Urbanization (UB)		Urban population growth (annual %)	[34]
	Economic Risk (ER)		Index	[37]

Table 2. Data and measurement.

Role	Variables	Dimension	Index Interpretation	Sources
Dependent Variables	Explicit Energy Transition (EET)	Energy–Environment	Ratio of non-fossil generation to total generation	[33]
		Energy–Environment-Society	Share of population with clean cooking fuels
		Energy–Environment–Technology	Share of GDP spent on R&D
		Energy–Environment–Technology	Patent-based measure of clean energy innovation
	Implicit Energy Transition (IET)	Energy–Environment–Society	Average energy consumption per capita	[33]
		Energy–Environment–Economy	Energy use per unit of GDP
		Energy–Environment–Politics	Electricity imports/consumption ratio
		Energy–Environment–Politics	Modern renewables’ share in final consumption
		Energy–Environment–Economy	CO2 emissions per unit of energy supply
Moderating Variables	Knowledge Production (KP)		Number of papers published, the number of papers cited, and the citation count	[33]
	Financial Development (FD)		Domestic credit to private sector by banks (% of GDP)	[34]
Control Variables	Human Capital (HC)		Index	[35]
	Economic Growth (EG)		GDP per capita (constant 2015 US$)	[34]
	Financial Globalization (FIG)		Index	[36]
	Technological Innovations (TI)		Patents Resident	[33]
	Urbanization (UB)		Urban population growth (annual %)	[34]
	Economic Risk (ER)		Index	[37]

3.2. Empirical Model

3.2.1. Benchmark Regression Model

To quantify the influence of AI on both explicit (EET) and implicit (IET) energy transitions, we estimate the following two-way fixed-effects panel model:

$${EET}_{i,t}=\alpha +\beta A{I}_{i,t}+{\gamma }^{\prime }{\mathrm{X}}_{i,t}+{\mu }_i+{\lambda }_t+{\epsilon }_{i,t},$$

(1)

$${IET}_{i,t}=\alpha +\beta A{I}_{i,t}+{\gamma }^{\prime }{\mathrm{X}}_{i,t}+{\mu }_i+{\lambda }_t+{\epsilon }_{i,t},$$

(2)

where ${EET}_{i,t}$ and ${IET}_{i,t}$ represent explicit energy transition (EET) or implicit energy transition (IET) for country $i$ in year $t$, respectively. $A{I}_{i,t}$ quantifies the intensity of artificial intelligence applications. ${\mathrm{X}}_{i,t}$ is the vector of control variables (human capital, economic growth, financial globalization, technological innovation, and urbanization). α is a constant term. β captures the direct effect of AI on energy transition. γ is a vector of coefficients on controls. μi and λt denote country and year fixed effects, respectively, absorbing time-invariant heterogeneity and common shocks and ${\epsilon }_{i,t}$ is the idiosyncratic error term.

To capture both explicit and implicit transition spillovers for country, we then estimate:

$${EET}_{i,t}={\alpha }_0+{\alpha }_1{\mathrm{A}\mathrm{I}}_{i,t}+{\alpha }_2^{\prime }{\mathrm{X}}_{j,t}+{\mu }_j+{\lambda }_t+{\epsilon }_{j,t},$$

(3)

$${IET}_{i,t}={\alpha }_0+{\alpha }_1{\mathrm{A}\mathrm{I}}_{i,t}+{\alpha }_2^{\prime }{\mathrm{X}}_{j,t}+{\mu }_j+{\lambda }_t+{\epsilon }_{j,t},$$

(4)

${\mathrm{A}\mathrm{I}}_{i,t}$ is the average AI intensity in the peer set $i\ne j$; ${\mathrm{X}}_{j,t}$ contains the controls (HC, EG, FIG, TI, UB) for country j; ${\mu }_j \mathrm{a}\mathrm{n}\mathrm{d} {\lambda }_t$ are country and year fixed effects; and ${\epsilon }_{j,t}$ is the error term. The spillover variable is constructed as the mean of that indicator across all BRICS countries except the one under study.

3.2.2. Mediating Effect Model

Beyond its direct impact on EET, AI can also affect EET indirectly through IET, a pathway we capture using a mediation-effect model:

$$IE{T}_{i,t}={\gamma }_0+{\gamma }_{1}A{I}_{i,t}+{\gamma }_2^{\prime }{\mathrm{X}}_{i,t}+{\mu }_i+{\lambda }_t+{\epsilon }_{i,t}$$

(5)

$$EE{T}_{i,t}={\delta }_0+{\delta }_{1}A{I}_{i,t}+{\delta }_{2}IE{T}_{i,t}+{\delta }_3^{\prime }{\mathrm{X}}_{i,t}+{\mu }_i+{\lambda }_t+{\epsilon }_{i,t}$$

(6)

In Equations (5) and (6), $IE{T}_{i,t}$ and $EE{T}_{i,t}$ function as mediators.

Building on the mediation-effect framework, we specify the two equations as follows:

$${\mathrm{I}\mathrm{E}\mathrm{T}}_{j,t}={\gamma }_0+{\gamma }_1{\mathrm{A}\mathrm{I}}_{i,t}+{\gamma }_2^{\prime }{\mathrm{X}}_{j,t}+{\mu }_j+{\sigma }_t+{\epsilon }_{j,t},$$

(7)

$${\mathrm{E}\mathrm{E}\mathrm{T}}_{j,t}={\delta }_0+{\delta }_1{\mathrm{A}\mathrm{I}}_{i,t}+{\delta }_2{\mathrm{I}\mathrm{E}\mathrm{T}}_{j,t}+{\delta }_3^{\prime }{\mathrm{X}}_{j,t}+{\mu }_j+{\sigma }_t+{\epsilon }_{i,t},$$

(8)

In Equations (7) and (8), ${\mathrm{I}\mathrm{E}\mathrm{T}}_{j,t}$ functions as the mediator, while all other model parameters and specifications remain identical to those in Equation (4).

3.2.3. Moderating Effect

Drawing on Equations (1) and (2), this study incorporates FD and KP as moderating factors: Equations (9) and (10) introduce interaction terms with FD, while Equations (11) and (12) incorporate interaction terms with KP.

$${\mathrm{E}\mathrm{E}\mathrm{T}}_{\mathrm{i}\mathrm{t}}={\alpha }_0+{\alpha }_{1}A{I}_{\mathrm{i}\mathrm{t}}+{\alpha }_{2}A{I}_{\mathrm{i}\mathrm{t}}^2+{\alpha }_{3}NR{D}_{\mathrm{i}\mathrm{t}}+{\alpha }_{4}A{I}_{\mathrm{i}\mathrm{t}}\ast F{D}_{\mathrm{i}\mathrm{t}}+{\alpha }_{5}A{I}_{\mathrm{i}\mathrm{t}}^2\ast NR{D}_{\mathrm{i}\mathrm{t}}+{\alpha }_6 \mathrm{Controls} +{\mu }_{\mathrm{i}}+{\sigma }_t+{\epsilon }_{\mathrm{i}\mathrm{t}}$$

(9)

$${\mathrm{I}\mathrm{E}\mathrm{T}}_{\mathrm{i}\mathrm{t}}={\beta }_0+{\beta }_{1}A{I}_{\mathrm{i}\mathrm{t}}+{\beta }_{2}A{I}_{\mathrm{i}\mathrm{t}}^2+{\beta }_{3}NR{D}_{\mathrm{i}\mathrm{t}}+{\beta }_{4}A{I}_{\mathrm{i}\mathrm{t}}\ast F{D}_{\mathrm{i}\mathrm{t}}+{\beta }_{5}A{I}_{\mathrm{i}\mathrm{t}}^{2\ast }NR{D}_{\mathrm{i}\mathrm{t}}+{\beta }_6 \mathrm{Controls} +{\mu }_{\mathrm{i}}+{\sigma }_t+{\epsilon }_{\mathrm{i}\mathrm{t}}$$

(10)

$${\mathrm{E}\mathrm{E}\mathrm{T}}_{\mathrm{i}\mathrm{t}}={\alpha }_0+{\alpha }_{1}A{I}_{\mathrm{i}\mathrm{t}}+{\alpha }_{2}A{I}_{\mathrm{i}\mathrm{t}}^2+{\alpha }_{3}K{P}_{\mathrm{i}\mathrm{t}}+{\alpha }_{4}A{I}_{\mathrm{i}\mathrm{t}}\ast K{P}_{\mathrm{i}\mathrm{t}}+{\alpha }_{5}A{I}_{\mathrm{i}\mathrm{t}}^2\ast K{P}_{\mathrm{i}\mathrm{t}}+{\alpha }_6{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}}_{\mathrm{i}\mathrm{t}}+{\mu }_{\mathrm{i}}+{\sigma }_t+{\epsilon }_{\mathrm{i}\mathrm{t}}$$

(11)

$${\mathrm{I}\mathrm{E}\mathrm{T}}_{\mathrm{i}\mathrm{t}}={\beta }_0+{\beta }_{1}A{I}_{\mathrm{i}\mathrm{t}}+{\beta }_{2}A{I}_{\mathrm{i}\mathrm{t}}^2+{\beta }_{3}K{P}_{\mathrm{i}\mathrm{t}}+{\beta }_{4}A{I}_{\mathrm{i}\mathrm{t}}\ast K{P}_{\mathrm{i}\mathrm{t}}+{\beta }_{5}A{I}_{\mathrm{i}\mathrm{t}}^2\ast K{P}_{\mathrm{i}\mathrm{t}}+{\beta }_6 \mathrm{Controls} +{\mu }_{\mathrm{i}}+{\sigma }_t+{\epsilon }_{\mathrm{i}\mathrm{t}}$$

(12)

4. Results

4.1. Descriptive Statistics

Table 3. Descriptive statistics.

Variable	Obs	Mean	Std. Dev.	Min	Max
EET	80	0.343	0.272	0.059	0.919
IET	80	0.299	0.134	0.130	0.705
AI	80	10.084	1.858	6.023	14.857
FD	80	81.861	41.108	25.923	179.104
KP	80	0.177	0.238	0.027	0.979
EG	80	8.578	0.721	6.854	9.266
UB	80	1.804	1.070	−0.320	3.882
HC	80	2.636	0.452	1.857	3.434
FIG	80	50.195	6.602	36.747	62.090
TI	80	9.511	2.203	6.295	14.148

summarizes the key features of studied variables. EET averages 0.343 (SD = 0.272) and ranges from 0.059 to 0.919, indicating substantial cross-country or cross-period variation in uptake. IET shows a lower mean of 0.299 with less dispersion (SD = 0.134), reflecting more consistency but still a wide span between its minimum (0.130) and maximum (0.705). Artificial intelligence intensity (AI) has a mean of 10.084 and a standard deviation of 1.858, with values stretching from 6.023 to 14.857, suggesting that some units are adopting AI at much higher levels than others. FD exhibits considerable heterogeneity (mean = 81.861, SD = 41.108), with low-end observations around 25.923 and high-end reaching nearly 180. KP average 0.177 but with a large coefficient of variation (SD = 0.238; min = 0.027, max = 0.979), highlighting that transformational innovation is unevenly distributed. EG is relatively stable (mean = 8.578, SD = 0.721), spanning 6.854 to 9.266 percent, while UB vary from –0.320 to 3.882 with a mean of 1.804 (SD = 1.070), indicating divergent social safety net generosity. Human capital (HC) scores average 2.636 (SD = 0.452) across a 1.857–3.434 range, pointing to modest variation in skill endowments. FIG centers at 50.195 (SD = 6.602), with most values between approximately 37 and 62, and TI averages 9.511 (SD = 2.203), ranging from 6.295 to 14.148.

Table 4. Correlation results.

	EET	IET	AI	FD	KP	EG	UB	HC	FIG	TI
EET	1	0.5769	0.0495	−0.5271	−0.1207	0.3449	−0.4668	0.1665	−0.1594	−0.0871
IET	0.5769	1	0.6369	−0.6493	0.6696	−0.1937	−0.3053	−0.052	−0.5088	0.0124
AI	0.0495	0.6369	1	−0.1173	0.8708	−0.5642	0.3526	−0.4796	−0.7204	0.1447
FD	−0.5271	−0.6493	−0.1173	1	−0.2736	0.2551	0.5559	−0.0844	0.2204	0.214
KP	−0.1207	0.6696	0.8708	−0.2736	1	−0.7356	0.3158	−0.5456	−0.6997	0.1187
EG	0.3449	−0.1937	−0.5642	0.2551	−0.7356	1	−0.5184	0.7749	0.6828	0.1397
UB	−0.4668	−0.3053	0.3526	0.5559	0.3158	−0.5184	1	−0.7507	−0.3935	0.1984
HC	0.1665	−0.052	−0.4796	−0.0844	−0.5456	0.7749	−0.7507	1	0.6822	0.0931
FIG	−0.1594	−0.5088	−0.7204	0.2204	−0.6997	0.6828	−0.3935	0.6822	1	0.172
TI	−0.0871	0.0124	0.1447	0.214	0.1187	0.1397	0.1984	0.0931	0.172	1

presents the pairwise Pearson correlations among the variables. EET and IET measures are strongly positively correlated (r = 0.5769), indicating that economies pursuing one form of energy transition often pursue the other in tandem. AI and KP exhibit an exceptionally high correlation (r = 0.8708), suggesting that AI adoption is closely linked to innovative output. FD is negatively associated with both implicit (r = −0.6493) and explicit (r = −0.5271) energy transition, implying that more advanced financial systems do not necessarily coincide with greener energy shifts. EG correlates strongly with HC (r = 0.7749) and moderately with FIG (r = 0.6828), underscoring the role of skills and cross-border financial integration in driving GDP growth. UB shows a moderate positive relationship with FD (r = 0.5559) but a marked negative correlation with HC (r = −0.7507), highlighting potential trade-offs between city expansion and workforce education levels. Other notable associations include AI–IET (r = 0.6369) and EG–KP (r = −0.7356), reflecting the complex interplay between technology, growth, and innovation.

4.2. Baseline Regression Results

The study examines the drivers of energy transition in BRICS nations.

Table 5. General regression result.

	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6	Model 7
Variable	EET	IET	EET	EET	IET	EET	IET
AI	0.0657 *** (4.4029)	0.1987 *** (7.0553)	−0.0306 ** (−2.2832)	0.0422 ** (2.0470)	0.1048 ** (2.9968)	−0.0248 * (−1.9698)	0.0098 (1.3572)
EG	0.0423 (0.4308)	0.3103 ** (2.6209)	−0.1081 (−1.3047)	0.0694 (0.6101)	0.5518 *** (3.8690)	−0.1091 (−1.1392)	0.0402 (1.0102)
UB	0.0085 (0.4200)	0.0149 (0.5677)	0.0013 (0.1067)	−0.0156 (−1.1626)	−0.0228 (−0.9343)	−0.0003 (−0.0257)	0.0040 (0.4949)
HC	0.0856 (1.3470)	0.5105 *** (7.0642)	−0.1619 ** (−2.5219)	0.0093 (0.1544)	0.4459 *** (5.4718)	−0.1374 * (−1.7492)	0.0991 ** (2.7293)
FIG	0.0048 ** (2.9226)	−0.0023 (−0.9928)	0.0059 *** (5.1501)	0.0047 ** (2.7911)	0.0001 (0.0627)	0.0065 *** (5.1380)	0.0015 ** (2.1131)
TI	−0.0188 (−0.6543)	−0.0766 ** (−2.5691)	0.0183 (0.7224)	−0.0014 (−0.0515)	−0.0657 ** (−2.1159)	0.0194 (0.6803)	−0.0078 (−0.6440)
IET			0.4848 *** (8.2962)
FD			−0.0021 *** (−5.4684)		−0.0035 *** (−4.8816)
AI*FD				−0.0000 (−0.1695)		−0.0005 ** (−2.4665)
KP						0.0700 (0.5086)	0.3468 *** (3.6825)
AI*KP						0.0433 * (1.7310)	0.0570 *** (3.7030)
CONS	−0.9840 (−1.4012)	−4.8967 *** (−6.1669)	1.3898 * (1.9867)	−0.7243 (−0.9208)	−5.7250 *** (−6.0347)	1.3529 (1.6570)	−0.4961 (−1.4397)
N	80	80	80	80	80	80	80
R2	0.99	0.94	0.99	0.99	0.96	0.99	0.99
id FE	YES	YES	YES	YES	YES	YES	YES
year FE	YES	YES	YES	YES	YES	YES	YES

Note: *** p < 1%, ** p < 5% and * p < 10%.

presents the baseline regression results. The baseline specifications (Models 1 and 7) exhibit excellent goodness-of-fit, with R² values of 0.99 for explicit energy transition (EET) and 0.94 for implicit energy transition (IET), respectively, and both include country and year fixed effects to absorb unobserved heterogeneity. The high explanatory power suggests that our chosen covariates collectively capture the main drivers of energy transitions in the BRICS bloc.

Model 1 reveals that AI has a positive and highly significant effect on explicit energy transition (EET) (0.0657, p < 1%). This indicates that artificial intelligence technologies directly support the shift from fossil fuels to renewables by improving energy system management, grid integration, and energy efficiency. The finding is consistent with [38], whose authors argue that AI-based applications such as smart grids and demand forecasting accelerate the penetration of renewable energy. Similarly, the authors of [6] highlight the transformative role of AI in clean energy innovation, strengthening the case that AI contributes positively to explicit transition outcomes.

Model 2 (IET as dependent variable) shows that AI strongly promotes implicit energy transition (IET) (0.1987, p < 1%), which refers to indirect improvements in energy efficiency, cleaner production processes, and broader digitalization of economic activities. The strength of this relationship suggests that AI’s influence is even more profound on the structural and systemic level than on direct renewable deployment. Studies such as [6,8] confirm that AI fosters energy-saving production methods, enhances industrial energy efficiency, and indirectly accelerates the path toward sustainability.

In Model 3 (EET with IET included), IET positively and significantly drives EET (0.4848, p < 1%), demonstrating a strong spillover effect. Implicit improvements in efficiency and digitalization ultimately translate into more explicit progress in renewable adoption. Interestingly, the coefficient of AI turns negative (−0.0306, p < 5%), suggesting that once IET is accounted for, much of AI’s direct effect on EET is mediated through IET. This aligns with evidence from [39], whose authors argue that digitalization and efficiency gains act as transmission mechanisms through which AI indirectly boosts renewable transitions.

Model 4 (EET with Financial Development) incorporates financial development (FD) alongside AI. The results show that FD negatively affects EET (−0.0021, p < 1%), indicating that higher financial development may still be biased toward fossil-fuel investments or traditional sectors rather than clean energy. The AI coefficient remains positive (0.0422, p < 5%), meaning AI independently drives EET, even in financially advanced economies. This finding resonates with [22], whose authors note that financial development often favors short-term profitability and carbon-intensive sectors, limiting its positive contribution to renewable transitions.

Model 5 (IET with Financial Development) examines IET with FD and AI included. Here, FD again exerts a negative and significant influence on IET (−0.0035, p < 1%), consistent with the argument that traditional financial systems do not adequately channel resources into efficiency-enhancing or sustainable practices. Conversely, AI positively affects IET (0.1048, p < 5%), reaffirming its role in indirect energy transformations. This dynamic suggests a tension between digital innovation and financial structures: while AI accelerates systemic change, financial development tends to reinforce established, carbon-intensive pathways. The authors of [17,40] similarly find that financial systems in emerging economies often lag in supporting sustainability transitions.

Model 6 (EET with Knowledge Production) introduces knowledge production (KP) and its interaction with AI. The coefficient of KP is insignificant (0.0700), but the interaction term AI*KP is positive and significant (0.0433, p < 10%). This indicates that knowledge production alone may not drive EET, but when combined with AI, it creates synergy that enhances explicit transition. In other words, AI becomes more effective in advancing EET in contexts with higher knowledge capacity. This result is supported by [41], whose authors argue that innovation and digital capacity must complement each other to produce substantial energy transition outcomes.

Model 7 (IET with Knowledge Production) shows that KP significantly promotes IET (0.3468, p < 1%), highlighting the role of research, patents, and knowledge intensity in driving systemic, indirect transitions. The interaction term AI*KP is also strongly positive (0.0570, p < 1%), suggesting a reinforcing effect: AI and knowledge production together accelerate implicit transitions more effectively than either alone. This reflects the argument of [42], whose authors stress that knowledge spillovers and AI-enhanced innovation are critical for sustaining long-term clean energy pathways.

4.3. Robustness Check

Table 6. Robustness test results.

	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6
	Replace and Add Control Variables			AI and Controls Lag One Period
Variable	EET	IET	EET	EET	IET	EET
AI	0.1131 *** (4.5493)	0.2051 *** (5.3548)	0.0179 (1.1006)
EG	−0.2819 * (−1.7091)	0.1470 (0.7741)	−0.3501 ** (−2.5473)
UB1	0.0123 ** (2.1487)	−0.0003 (−0.0449)	0.0124 ** (2.7065)
HC	0.0336 (0.4870)	0.4958 *** (5.1027)	−0.1965 ** (−2.7742)
FIG	0.0029 (1.4042)	−0.0016 (−0.6538)	0.0036 ** (2.2586)
TI	0.0004 (0.0138)	−0.0490 (−1.6125)	0.0231 (1.0777)
ER	0.0056 ** (3.1489)	0.0073 ** (2.2606)	0.0022 ** (2.3810)
IET			0.4641 *** (8.7953)			0.4698 *** (7.7861)
L.AI				0.0488 ** (2.6748)	0.1980 *** (5.3693)	−0.0442 ** (−2.7756)
L.EG				0.0585 (0.5112)	0.3855 ** (2.7542)	−0.1226 (−1.2589)
L.UB				0.0130 (0.5986)	0.0134 (0.4392)	0.0067 (0.4620)
L.HC				−0.0033 (−0.0394)	0.5276 *** (5.0661)	−0.2512 ** (−3.4074)
L.FIG				0.0021 (1.3959)	−0.0022 (−0.8523)	0.0032 ** (2.3910)
L.TI				−0.0222 (−0.6947)	−0.0905 ** (−2.5607)	0.0203 (0.7079)
CONS	0.4269	−4.0364 ***	2.3004 **	−0.5573	−5.4164 ***	1.9873 **
N	80	80	80	75	75	75
R2	0.9948	0.94	0.99	0.99	0.93	0.99
id FE	YES	YES	YES	YES	YES	YES
year FE	YES	YES	YES	YES	YES	YES

Note: *** p < 1%, ** p < 5% and * p < 10%.

provides robustness tests for the baseline results in Table 5 by introducing two strategies: (i) replacing and adding new control variables such as environmental regulation (ER), and (ii) lagging the main explanatory variable (AI) and controls by one period to reduce simultaneity and endogeneity concerns. These checks are essential to verify whether the observed relationships between AI, explicit energy transition (EET), and implicit energy transition (IET) remain stable when alternative specifications are applied.

Models 1–3 (alternative controls added) confirm the positive role of AI in both explicit and implicit transitions. AI significantly promotes EET (0.1131, p < 1%) and IET (0.2051, p < 1%), consistent with Table 5’s findings. This reinforces the conclusion that AI drives both direct renewable adoption and indirect systemic efficiency gains. Furthermore, the newly introduced control variable ER (economic risk) is positive and significant across models, implying that low economic risk strengthens the transition process. The persistence of AI’s positive effect, even after adjusting the control set, indicates strong robustness.

In Model 3 (Mediation via IET), where IET is included as an explanatory variable, the results again show that IET significantly promotes EET (0.4641, p < 1%), closely matching Table 5 (0.4848, p < 1%). The stability of this relationship across both baseline and robustness models indicates that implicit transitions—such as efficiency improvements and digital integration—are a crucial transmission channel through which AI exerts influence on explicit transitions. This robustness underscores the argument of [38], who highlight the mediating role of efficiency-led transitions in renewable expansion.

The lagged regressions (Models 4–6) provide an additional robustness check by addressing possible reverse causality. Lagged AI (L.AI) remains positive and significant for both EET (0.0488, p < 5%) and IET (0.1980, p < 1%), supporting the causal interpretation that AI adoption precedes and facilitates energy transition. Interestingly, when IET is added as a mediator, lagged AI turns negative (−0.0442, p < 5%), mirroring the pattern in Table 5 where AI’s direct effect on EET weakened once IET was controlled. This further validates the conclusion that much of AI’s impact on EET operates indirectly through IET, confirming the mediation mechanism.

Taken together, Table 6 demonstrates that the key findings in Table 5 are robust to alternative controls and lagged specifications. AI consistently enhances both explicit and implicit energy transitions, while implicit transition remains a strong driver of explicit transition. The inclusion of environmental regulation strengthens the model, suggesting institutional quality matters. Lagged results confirm the temporal precedence of AI in promoting transition, mitigating endogeneity concerns. Thus, the robustness tests reinforce the credibility of the main results, aligning with prior evidence that AI and digital innovations are critical enablers of sustainable energy pathways [4,7].

4.4. Spillover Effects of Artificial Intelligence (AI)

Considering cross-border influences under the BRICS cooperation framework,

Table 7. Spillover effect results.

	Model 1	Model 2	Model 3
Variable	EETj	IETj	EETj
${AI}_j$	−0.0160 *** (−4.6363)	−0.0511 *** (−6.8569)	−0.0064 (−1.3588)
EG	−0.0513 ** (−2.5963)	−0.0871 ** (−2.7688)	−0.0351 * (−1.7040)
UB	−0.0060 (−1.3692)	−0.0053 (−0.7457)	−0.0050 (−1.1884)
HC	−0.0270 (−1.3758)	−0.1317 *** (−6.9579)	−0.0024 (−0.1223)
FIG	−0.0006 (−1.4078)	0.0005 (0.8832)	−0.0007 * (−1.6906)
TI	0.0140 ** (2.8305)	0.0205 ** (2.6087)	0.0102 * (1.9368)
${IET}_j$			0.1868 ** (2.9069)
CONS	0.7367 *** (4.5581)	1.6900 *** (8.0575)	0.4209 ** (2.2382)
N	80	80	80
R2	0.998	0.9772	0.9982
id FE	YES	YES	YES
year FE	YES	YES	YES

Note: t statistics in parentheses * p < 0.1, ** p < 0.05, *** p < 0.01.

presents three fixed-effects spillover regressions for explicit (EET_j) and implicit (IET_j) energy transitions. In Model 1, a one-unit increase in partner-country AI intensity (AI_j) significantly reduces domestic explicit transition (β = −0.0160, t = −4.64, p < 0.01), while partner economic growth (EG_j) likewise exerts a negative spillover (β = −0.0513, t = −2.60, p < 0.05). Urbanization (UB_j) and human capital (HC_j) spillovers are negative but insignificant, whereas financial globalization (FIG_j) also yields a marginal negative effect (β = −0.0006, t = −1.41, p > 0.10). By contrast, partner technological innovation (TI_j) produces a significant positive externality (β = 0.0140, t = 2.83, p < 0.05), suggesting that cross-border diffusion of clean-tech patents and R&D raises each country’s visible energy-mix shifts. The strong constant (0.7367, t = 4.56) and R² = 0.998 demonstrate that, net of neighbor effects, there remains a high baseline propensity for transition across BRICS.

Model 2, which targets implicit transition spillovers (IET_j), shows an even larger negative externality from AI_j (β = −0.0511, t = −6.86, p < 0.01) and EG_j (β = −0.0871, t = −2.77, p < 0.05). Notably, partner human capital exerts a strong negative drain on implicit improvements (β = −0.1317, t = −6.96, p < 0.01), indicating that talent migration toward leading BRICS economies hinders neighbors’ behind-the-meter efficiency gains. Urbanization and FIG spillovers remain statistically indistinct from zero, while TI_j again contributes positively to efficiency spillovers (β = 0.0205, t = 2.61, p < 0.05). The intercept of 1.6900 (t = 8.06) and R² = 0.9772 confirm robust baseline momentum in implicit transition, even as cross-border competition erodes some gains. In Model 3, we reintroduce partner implicit transition (IET_j) alongside AI_j to explain domestic explicit shifts. Here, the negative AI_j coefficient shrinks to insignificance (β = −0.0064, t = −1.36, p > 0.10), while IET_j emerges as a significant positive demonstrator (β = 0.1868, t = 2.91, p < 0.05). Partner economic growth continues to depress domestic explicit transition (β = −0.0351, t = −1.70, p < 0.10), and FIG_j retains a weak negative effect (β = −0.0007, t = −1.69, p < 0.10). Partner TI_j remains positively associated (β = 0.0102, t = 1.94, p < 0.10). These results indicate that while initial AI and growth competition siphons resources away, observing neighbors’ efficiency improvements ultimately spurs one’s own visible energy-mix transitions. The constant of 0.4209 (t = 2.24) and R² = 0.9982 attest to the model’s explanatory power.

Several factors help explain these patterns. First, uneven AI capabilities across BRICS—led by China and India with advanced AI ecosystems—draw investment and specialized talent, weakening smaller members’ transition efforts. Second, policy and technological barriers hamper seamless AI diffusion: disparate data-governance regimes and variable digital infrastructure slow technology transfer, limiting countries’ ability to translate peer advances into domestic gains [8]. Third, despite high-level dialogues within the BRICS framework, formal mechanisms for joint AI–energy R&D and resource pooling remain underdeveloped, delaying the positive demonstration effects of efficiency innovations. Empirical studies corroborate these spillovers. The authors of [7] found that AI-driven grid optimizations in China inadvertently reduce foreign green-tech investment, while the authors of [43] document how R&D spillovers in clean-tech boost neighboring India’s renewables share only after efficiency benchmarks are met. Meanwhile, the authors of [38] highlight that patent co-development agreements in the BRICS have yielded positive TI spillovers, aligning with our positive TI_j coefficients.

Policy implications are clear: to mitigate negative resource competition, BRICS should harmonize AI standards, establish joint green-innovation funds, and create talent-exchange platforms. Strengthening formal R&D consortia will allow lagging members to access peer efficiency breakthroughs directly, transforming initial “beggar-thy-neighbor” dynamics into sustained, mutual advancement in both explicit and implicit energy transitions.

5. Conclusions and Policy Recommendations

5.1. Conclusions

As the BRICS nations stand at the crossroads of a global energy revolution, harnessing AI’s dual engines of knowledge production and financial innovation has never been more critical. By aligning “smart money” with “smart minds,” these emerging powerhouses can turbocharge their transition to a low-carbon future. Thus, this study used the BRICS nations in examining the factors driving explicit structure and implicit order of energy transition from 2005 to 2020. The analysis proceeds through five key stages: baseline regression results establish that AI intensity is the primary driver of both visible (EET) and efficiency-based (IET) energy transitions in BRICS, with economic growth, human capital, and financial globalization playing secondary roles; mediation effects of IET reveal that AI’s impact on explicit transition is fully channeled through efficiency gains; moderating effects of financial development and knowledge production show that deep financial markets can crowd out AI’s green benefits while robust patenting ecosystems amplify its efficiency impact; robustness check results confirm the stability of these relationships across alternative controls, lag structures, and model specifications; and finally, spillover effects highlight how cross-border AI advances, growth trajectories, human capital flows, financial linkages, and innovation spillovers collectively shape each country’s own energy-transition trajectory within the BRICS cooperation framework.

5.2. Policy Recommendations

Based on our empirical findings across baseline, mediation, moderating, robustness, and spillover analyses, several policy implications emerge for BRICS nations:

(a)

Phase AI Integration through Efficiency Pathways: Given AI’s strong direct effects on both explicit and implicit energy transitions—and its full mediation via efficiency gains—governments should adopt a phased approach to AI deployment in energy systems. In the short term, pilot AI-enabled smartgrid and demand-response projects (e.g., China’s New Generation Smart Substation and India’s Green Energy Corridor) can demonstrate tangible efficiency improvements. Over the medium to long term, mandatory connections to a national AI-powered energy-management platform, coupled with dedicated R&D funding for grid optimization algorithms, will ensure AI investments translate into sustained renewable integration and loss reductions.

(b)

Strengthen Implicit Transition via Human Capital and Knowledge Production: Our results show that human capital and patenting intensity are critical for behind-the-meter improvements. BRICS governments should bolster technical and vocational training in energy technologies and incentivize clean-tech patenting through R&D tax credits and innovation vouchers. Establishing joint research centers and patent pools focused on energy-efficiency solutions will magnify AI’s impact on system optimization and drive the positive demonstration effects that, in turn, promote visible shifts in energy mixes.

(c)

Reform Financial Systems to Mobilize Green Capital: Deep financial markets have shown a crowding-out effect on green investments unless steered by policy. Mandating a minimum green-loan quota for commercial banks and expanding green bond markets will redirect capital from legacy fossil sectors toward renewables. Further, harmonizing BRICS-wide standards for sustainable finance—such as unified taxonomy and reporting rules—can leverage financial globalization to amplify explicit transition without diluting AI’s green potential.

(d)

Leverage Urbanization and Regulatory Stringency: Denser urban centers facilitate pilot green infrastructure and district-level efficiency upgrades. Urban planning should integrate renewables and microgrids into new developments, while environmental regulations—already shown to be robust levers—must be tightened to enforce higher efficiency benchmarks and renewable-energy quotas for utilities. Coordinated urban energy policies will harness demographic trends for rapid deployment of clean technologies.

(e)

Deepen Cross-Border Cooperation and Demonstration Spillovers: To counteract negative AI and growth spillovers, BRICS should create formal joint R&D consortia, shared data-governance protocols, and talent-exchange programs. Demonstration projects—such as cross-country pilot grids—can accelerate learning and reduce competitive resource drains. Aligning AI standards and IP regimes across the bloc will enable faster technology diffusion and turn initial “beggar-thy-neighbor” dynamics into sustained, bloc-wide advancement.

(f)

Differentiate Strategies by National Conditions: Countries with advanced AI infrastructure (China, India) should pioneer large-scale smart-grid demonstrations and open-source AI toolkits, while those with structural constraints (South Africa, Brazil) ought to focus first on off-grid renewables and microgrid interconnections to alleviate energy poverty. Developed members can support these tailored pathways through targeted grants, technical assistance, and co-investment vehicles.

(g)

Transition to Knowledge-Driven Green Development: Moving beyond resource dependence, governments must establish ecological red-lines and channel incentives toward low-carbon circular-economy projects. Prioritizing AI ethics and algorithmic transparency in energy allocation will ensure that smart energy systems serve all communities equitably. International green-technology platforms—facilitating real-time data sharing and joint patent licensing—can accelerate the shift toward knowledge-intensive, clean-energy industries.

5.3. Limitation and Future Direction

This study’s findings should be interpreted with four caveats that shape how the results are read. First, the sample is confined to the original five BRICS; as the bloc expands to BRICS-10, differences in institutions, energy mixes, and digital capacity may amplify or dampen the estimated AI–energy transition links, so external validity is limited to large emerging economies with similar structures. Second, our proxy for national AI—industrial robot adoption—primarily captures embodied automation in manufacturing and likely underrepresents data-centric AI in services, grids, and digital platforms; coefficients should therefore be read as partial (possibly lower-bound) effects. Future work should triangulate multiple indicators (e.g., ML patents, compute capacity, AI service deployments) and exploit firm- or project-level microdata to uncover heterogeneity in channels and sectors. Third, we do not model the social and ethical dimensions of AI adoption; algorithmic bias, data governance, and labor displacement can alter diffusion paths and energy equity, meaning our efficiency-oriented gains should not be construed as welfare-neutral. Incorporating fairness metrics and access outcomes would clarify who benefits and who may be left behind. Finally, while fixed-effects and spillover models illuminate dynamics, they do not fully resolve endogeneity; quasi-experimental designs (e.g., exogenous AI infrastructure rollouts, eligibility thresholds) and comparative analyses with advanced economies would strengthen causal identification and test the generalizability of our results.