Bridging the Digital–Energy Divide: Artificial Intelligence, Internet Connectivity, and Knowledge Management

Abstract

Achieving sustainable growth in emerging economies requires more than expanding clean energy; it also relies on the synergistic role of Artificial Intelligence, Internet Connectivity, and Knowledge Management in narrowing the digital–energy divide. Thus, this study examines the factors influencing the energy transition—both implicit and explicit—using the case of the BRICS economies with data spanning from 2000 to 2022. This study employed Driscoll–Kraay (DK) standard errors together with Lewbel IV-2SLS estimators to examine the connections. The results showed that Artificial Intelligence and economic growth hinder energy transition, while financial development and trade openness promote it. Furthermore, Knowledge Management and Internet Connectivity show threshold effects, and education remains negatively aligned with sustainability goals. Based on these findings policies are proposed.

1. Introduction

Greenhouse gas (GHG) emissions continue to rise globally, with industry and transportation serving as primary sources. Fossil fuels remain central to meeting the energy demand, with petroleum, coal, and natural gas contributing about 36%, 27%, and 23.4%, respectively. Their combustion generates nearly 21.3 billion tons of CO₂ each year, making them a dominant greenhouse gas emitter and a critical driver of global warming [1].

The global imperative to mitigate climate change has thrust the energy transition to the forefront of policy, especially for BRICS economies that must sustain growth while cutting emissions and managing a large share of the global energy use. Clean options bring clear benefits: nuclear and hydrogen energy offer very high energy density and wide versatility across power, heat, and industry domains, while wind and solar power deliver low operating costs, modular deployment, fast build times, and improved security from domestic resources [2]. They also pose challenges, including high upfront capital needs, grid reinforcement and storage, charging or pipeline infrastructure, intermittency management for wind and solar, critical mineral supply risks, land and water constraints, local siting concerns, and lengthy permitting [3,4]. For BRICS, the most resilient pathway is a balanced portfolio that pairs a firm low-carbon capacity, such as nuclear and geothermal energy, with variable renewables, storage, demand responses, and expanded transmission, complemented by policies that lower financing costs and accelerate deployment, so that security, equity, and sustainability advance together.

The BRICS energy landscape is experiencing dynamic yet uneven progress. For instance, solar power generation across the group surged by 39% year-on-year in early 2025, with China leading at a 42% increase, followed by Brazil (35%) and India (32%) [5]. Hydropower still dominates renewable electricity shares at 45%, but solar PV and wind power have seen exponential growth—from 49 TWh to 792 TWh and from 245 TWh to 1089 TWh, respectively, since 2015. Investments have tracked this growth—China’s annual clean energy investment soared from USD 211 billion in 2019 to USD 818 billion in 2024, while India and Brazil mobilized about USD 47 billion and USD 39 billion, respectively (https://zerocarbon-analytics.org/archives/netzero/renewables-bulletin-brics-edition/) (accessed on 4 August 2025). Yet, structural hurdles persist: the ongoing reliance on coal and fossil-based energy systems, a lack of cohesive transition strategies among member nations, and limited climate finance remain inhibitors to more equitable and robust progress [6].

Artificial Intelligence (AI) offers transformative potential in steering BRICS energy systems toward transition goals. AI can enhance predictive grid stability, forecast renewable generation, optimize energy storage usage, and reduce peak power loads, directly contributing to both Explicit and Implicit Energy Transitions [7,8]. In practical terms, AI-informed demand–response mechanisms and smart energy scheduling reduce waste and elevate grid resilience. However, deployment must be strategic: an overreliance on computational infrastructure may exacerbate the energy demand unless accompanied by clean power sources [9]. Thus, AI presents both significant promise and a cautionary tale—its benefits hinge on thoughtful integration with sustainable energy frameworks.

Knowledge Management (KM) systems serve as pivotal conduits for energy innovation, enabling knowledge flow, institutional learning, and policy evolution. Effective KM mobilizes R&D outputs, patent insights, and skilled human capital to accelerate renewable energy deployment and technological diffusion [10,11]. In the context of BRICS, robust KM platforms can bridge the gap between scientific discovery and policy execution, facilitating the localized adaptation of renewables and energy efficiency initiatives. By enhancing absorptive capacity, KM amplifies the efficacy of technological tools—such as AI and smart infrastructure—in driving sustainable transitions across diverse socio-economic landscapes.

Internet Connectivity (INTC) underpins the digital transformation integral to modern energy systems. High-speed broadband, mobile networks, and secure digital platforms enable real-time monitoring, decentralized energy management, and the integration of prosumer models [12]. In BRICS contexts where connectivity varies widely, enhancing digital infrastructure can unlock demand-side flexibility, smart metering, and scalable renewable integration [4]. Yet, this is not without risk: unchecked connectivity expansion may lead to rebound effects, where increased digital activity drives up energy consumption unless balanced by efficiency improvements and renewable sourcing [13]. Therefore, connectivity must be strategically harnessed to support—not hinder—the clean energy transition. Based on the above information, this study investigates the effects of Artificial Intelligence (AI) on both the Explicit and Implicit Energy Transition and the influence of Knowledge Management (KM) and international technology collaboration (INTC) on these transitions.

This study contributes to the growing body of literature on sustainability and digital transformation by offering a comprehensive framework that jointly examines Artificial Intelligence (AI), Knowledge Management (KM), and Internet Connectivity (INTC) as drivers of both Explicit and Implicit Energy Transitions. While prior research has investigated these elements separately, few studies integrate them within a single empirical framework in the context of BRICS. More importantly, to the best of our knowledge, this is the first study to test whether the Implicit Energy Transition (IET) mediates the relationship between AI and the Explicit Energy Transition (EET) and whether INTC and KM moderate the relationship between AI and energy transition (explicit and implicit). This novelty allows this study to provide a deeper understanding of the complex interaction mechanisms shaping clean energy pathways in emerging economies.

Second, this study makes a methodological contribution by applying advanced econometric techniques—including Driscoll–Kraay standard errors and Lewbel IV-2SLS estimators—to tackle potential endogeneity and nonlinearity. This dual approach strengthens causal inference and ensures the robustness of the results, particularly in analyzing complex dynamics such as the mediating role of IET and the moderating roles of KM and INTC. By empirically testing these new relationships for the first time, this study enhances methodological innovation and sets a precedent for future research investigating how digitalization and knowledge ecosystems intersect with sustainability transitions.

Finally, the findings provide practical contributions for policymakers and practitioners in the BRICS nations by highlighting actionable pathways to accelerate energy transition. Specifically, the results show that AI’s rebound effects can be mitigated through stronger KM systems, while INTC expansion must be carefully managed to avoid the digital energy demand outpacing efficiency gains. This evidence supports a policy design that emphasizes digital infrastructure aligned with renewable energy policies, institutional learning, and human capital development. Thus, this study not only advances academic debates but also offers evidence-based guidance for governments, investors, and energy stakeholders seeking to balance economic growth with climate commitments.

The remainder of this paper is structured as follows: Section 2 presents the theoretical framework and literature review, Section 3 outlines the data and methodology, Section 4 reports the empirical findings, and Section 5 concludes with policy recommendations.

2. Theoretical Framework and Empirical Literature

2.1. Theoretical Framework

Theoretically, the relationship between Artificial Intelligence (AI), Internet Connectivity (INTC), and Knowledge Management (KM) with energy transition can be explained through the lenses of innovation diffusion, resource optimization, and system efficiency. AI, as a general-purpose technology, facilitates predictive analytics, smart grid optimization, and renewable energy integration, thereby directly advancing the Explicit Energy Transition (shifts in the energy mix toward renewables) while also improving implicit transitions through efficiency gains [2,8]. However, AI’s impact is not linear; in early stages, it may increase the energy demand due to computational loads, but at more advanced levels, it reduces emissions by optimizing processes and enabling the large-scale integration of clean energy [9].

Similarly, Internet Connectivity acts as a critical enabler by providing the digital infrastructure for real-time monitoring, decentralized energy management, and cross-border technology transfer [14]. Broadband and mobile diffusion underpins smart metering, demand responses, and virtual power plants, while empowering prosumers to generate, store, and trade electricity via platform-based markets [12]. High-quality connectivity also lowers coordination costs for integrating variable renewables, supports predictive maintenance through IoT and edge analytics, and facilitates interoperability via common data standards. When paired with robust cybersecurity and inclusive access (urban and rural), these capabilities accelerate the pace and breadth of the energy transition [15], though excessive connectivity may create rebound effects when a rising digital energy demand outpaces efficiency gains [13]. Knowledge Management (KM) complements this process by fostering innovation ecosystems, strengthening absorptive capacity, and enabling sustainable technology deployment at scale. Effective KM architectures—combining repositories and taxonomies, communities of practice, after-action reviews, and standardized playbooks—convert tacit and codified knowledge from R&D, patents, and human capital investments into deployable green solutions [16]. By structuring flows between universities, firms, and regulators, KM shortens learning cycles, diffuses best practices across sites, and reduces the duplication and integration risk. It also aligns incentives for knowledge sharing, safeguards IP while enabling interoperability, and embeds continuous training and feedback loops for operations and maintenance [17]. Theoretically, KM provides the learning mechanisms that transform technological inputs into systemic change, allowing AI-driven insights and digitally enabled energy solutions to diffuse effectively across industries and societies.

2.2. Empirical Review

Artificial Intelligence (AI) emerges as a catalytic—but contingent—driver of the energy transition. Dynamic macro-panel evidence shows that AI lowers ecological footprints and emissions and accelerates transition once enabling thresholds, such as trade openness, are crossed [2], while a 50-country panel confirms a broad, statistically significant association between AI diffusion and transition progress [15]. These gains materialize through concrete operational channels: AI improves forecasting, unit commitment, and storage dispatch, thereby easing solar/wind integration [18]; enhances market resilience and demand-side flexibility when price signals and data access are well designed [8]; and even amplifies nuclear power’s abatement effect, particularly in higher-emitting strata (Danish & Khan, 2025 [9]). Crucially, the benefits are governance-conditional: CS-ARDL estimates indicate that AI supports energy transition only where the regulatory capacity and rule of law govern deployment (Işık et al., 2024 [17]), consistent with U.S. evidence that AI mitigates environmental stress under coherent regulation [13]. The literature therefore converges on a clear argument: AI is an accelerant that converts good institutions, market rules, and complementary assets (e.g., nuclear, storage, and trade) into measurable decarbonization.

Table 1. Impact of Artificial Intelligence on energy transition.

Author(s)	Focus/Title	Methodology	Key Insight
[2]	AI’s multifaceted impact on sustainability	SYS-GMM, threshold panel	AI reduces ecological footprints and emissions; boosts transition especially with trade
[17]	AI + governance in ET	CS-ARDL	AI helps only when governance is strong
[18]	AI for renewable integration management	Modeling review	AI speeds adoption of solar/wind renewables
[15]	AI and energy transition	Panel study of 50 countries	AI significantly impacts transition across nations
[8]	AI’s role in energy markets	Modeling + demand-side strategies	Enhances market resilience and renewable use
[9]	AI + nuclear CO2 effects	Quantile panel regression	AI amplifies nuclear power’s emission reduction
[13]	AI and green economy in the US	ARDL STIRPAT	AI mitigates environmental stress when regulated

presents the summary of the impact of Artificial Intelligence on the energy transition.

Knowledge Management (KM) functions as the organizational backbone that translates technologies and policies into a sustained transition capability. Case and interview evidence shows that smart KM systems and KM practices are pivotal in firm-level transformation and national renewable roll-outs [19]. Conceptual and empirical work links KM to the energy culture and learning—codifying best practices, diffusing tacit know-how, and reinforcing pro-efficiency norms that lower the demand and hasten the clean technology uptake [11,16,20]. Public information ecosystems matter too: a survey analysis indicates that media exposure enlarges the stock of transition knowledge among citizens, easing the acceptance of new tariffs, appliances, and mobility choices [21]. Collectively, these studies argue that KM is not an accessory but a core production factor for the transition, raising absorptive capacity, shortening learning curves, and aligning behavior across firms, workers, and households.

Table 2. Impact of Knowledge Management on energy transition.

Author(s)	Focus/Title	Methodology	Key Insight
[19]	Smart KM systems	Case study	KM accelerates energy firm transformation
[10]	KM in Vietnam	Interviews	KM critical to renewable deployment
[11]	KM and energy culture	KM model	Promotes energy-saving behaviors
[20]	Learning and clean energy	Empirical learning analysis	Learning supports energy transitions
[16]	KM performance in RET	Empirical study	KM enhances RE technology uptake
[21]	Media’s role in Transition knowledge	Survey analysis	Media boosts public energy transition knowledge

presents the summary of the impact of Knowledge Management on the energy transition

Internet Connectivity provides the infrastructure for real-time information, coordination, and control that modern low-carbon systems require. Empirical models find that internet development directly supports renewables’ deployment [12] and that rural connectivity increases household shifts to advanced, cleaner fuels [4]. At a system scale, “energy internet” architectures accelerate the clean energy supply by enabling distributed optimization and two-way flows [22], while applied models highlight the interactive role of data centers and grid digitalization in shaping transition pathways (Wan et al., 2023 [23]). Yet heterogeneity is salient: the uptake of home energy management systems depends on digital competence, with the digital divide constraining benefits [24]. Consistent panel evidence shows that broader digital transformation both mitigates emissions linked to the transition and supports growth when embedded in sound policy [2,25]. The argument is thus two-handed: connectivity unlocks efficiency, flexibility, and adoption at scale, provided the capability gaps and rebound from digital loads are managed.

Table 3. Impact of Internet Connectivity on energy transition.

Author(s)	Focus/Title	Methodology	Key Insight
[12]	Internet development on RETI	Empirical models	Internet significantly supports renewables
[4]	Rural internet use	Ordered probit model	Increases rural move to advanced fuels
[24]	Digital divide in ET	Social analysis	HEMS use linked to digital competency
[22]	Energy internet effect	Scenario modeling	Accelerates clean energy supply
[23]	Data centers and electric transition	Applied energy model	Interactive role in network transition
[2]	China digitization and CO2	Regression analysis	Digitization mitigates emissions tied to transition
[25]	Digital transformation and clean energy	Panel ARDL	Digital growth supports transition and growth

presents the summary of the impact of Internet Connectivity on the energy transition

Economic growth exerts an asymmetric, often nonlinear influence on the transition, implying that policy design—not growth per se—determines decarbonization outcomes. Multi-country CS-ARDL results suggest that growth on its own can impede the transition—through scale and composition effects—unless counterbalanced by governance and sustainability constraints, including AI and ESG frameworks [26]. At the same time, evidence of nonlinearity shows that a well-managed transition can yield growth dividends, with spline and QQR models identifying thresholds and quantile-specific effects in which renewable expansion and efficiency raise the output while high emissions growth hinders progress [27,28]. Macroeconomic fundamentals—prices, investment, and policy credibility—shape the renewable shift [29,30], and reviews indicate that, within sustainability bounds, renewables support long-run growth [31]. Digitalization can tilt this balance by amplifying the positive growth effects of transition through productivity and coordination gains [32]. Together, the literature argues for sequencing: use digital and institutional complements to bend growth onto a low-carbon path, recognizing heterogeneous responses across the distribution of economies and over the development cycle.

Table 4. Impact of economic growth on energy transition.

Author(s)	Focus/Title	Methodology	Key Insight
[26]	Growth, ESG, AI vs. ET	CS-ARDL	Economic growth alone can impede transition
[27]	ENT and Growth	Spline regression	Energy transition yields growth with nonlinearity
[3]	Macroeconomics on RE	Modeling regression	Economic factors shape renewable shift
[28]	GDP and ENT in UK	QQR modeling	Asymmetric GDP impact across entropy quantiles
[31]	RE and growth in BRICS/Brazil	Empirical chapter review	Renewable energy supports growth within sustainable bounds
[32]	Digitalization’s moderating role	Regression models	Digitization can amplify transition’s growth effect

presents the summary of the impact of economic growth on the energy transition.

2.3. Hypotheses Formulation

The hypotheses for this empirical investigation are formulated as follows to provide a clear direction for this study.

Direct Effects (Baseline Models)

H1a: 

Artificial Intelligence (AI) has a significant effect on the Explicit Energy Transition (EET).

H1b: 

Artificial Intelligence (AI) has a significant effect on the Implicit Energy Transition (IET).

H2: 

Economic growth (lnEG) negatively affects both the EET and IET due to fossil fuel dependency.

H3: 

Financial development (FD) positively influences the EET and IET by facilitating green investment.

H4: 

Trade openness (TR) positively influences the EET and IET by enhancing technology diffusion.

H5: 

Education (EDU) has a significant effect on the EET and IET, but its impact may be misaligned with sustainability goals.

Mediation Effect

H6: 

The Implicit Energy Transition (IET) mediates the relationship between AI and the Explicit Energy Transition (EET).

Moderating Effects (Linear Models)

H7: 

Knowledge Management (KM) moderates the relationship between AI and energy transition (EET and IET).

H8: 

Internet Connectivity (INTC) moderates the relationship between AI and energy transition (EET and IET).

Nonlinear (Inverted-U) Effects

H9: 

Knowledge Management (KM) has an inverted-U relationship with IET, strengthening transition up to a threshold but diminishing at higher levels.

H10: 

Internet Connectivity (INTC) has an inverted-U relationship with both the EET and IET, enhancing the transition at moderate levels but reversing at high penetration due to rebound effects.

3. Data and Methods

3.1. Data

This study examines the determinants of Explicit and Implicit Energy Transitions in BRICS economies between 2000 and 2022, drawing on a structured framework that integrates dependent, independent, mediating, moderating, and control variables. Explicit Energy Transition (EET) reflects observable shifts in the energy mix toward renewables, while Implicit Energy Transition (IET) captures efficiency-oriented improvements such as lower energy intensity. Artificial Intelligence (AI) is the core independent variable, expected to influence both transitions directly and indirectly, with IET functioning as a mediator that channels AI’s role in promoting system-level modernization into measurable renewable adoption. The framework further incorporates two moderating factors—Knowledge Management (KM) and Internet Connectivity (INTC)—to test whether institutionalized innovation and digital infrastructure amplify or condition the effect of AI, including potential nonlinear (inverted-U) dynamics. A set of controls—economic growth, financial development, trade openness, and education—accounts for macroeconomic and structural drivers that may either support or constrain transition. Together, this design enables a nuanced assessment of how AI, embedded within broader knowledge and connectivity ecosystems, interacts with contextual factors to shape the pace and trajectory of energy transitions in the BRICS region. All the data are sourced from the world development indicator (WDI) database.

Table 5. Data source and measurement.

Variable	Meaning	Abbreviation	Source	Measurement
Explicit Energy Transition	Observable shifts in the energy mix toward renewables	EET	[33]	Index
Implicit Energy Transition	Efficiency-oriented improvements such as lower energy intensity	IET	[33]	Index
Artificial Intelligence	Core independent variable expected to influence both transitions	AI	[33]	AI-related indicators (e.g., patents, ICT adoption, or proxy variables)
Knowledge Management	Moderating factor reflecting institutionalized innovation	KM	[33]	Knowledge economy or R&D indicators (% of GDP, researchers per million)
Internet Connectivity	Moderating factor capturing digital infrastructure	INTC	[33]	Internet users (% of population)
Economic Growth	Control variable reflecting macroeconomic expansion	EG	[33]	GDP per capita (constant 2015 USD)
Financial Development	Control variable for financial sector depth and efficiency	FD	[33]	Domestic credit to private sector (% of GDP)
Trade Openness	Control variable representing global integration	TO	[33]	Trade (% of GDP)
Education	Control variable capturing human capital	EDU	[33]	School enrollment or mean years of schooling

presents the detailed data information.

3.2. Empirical Model

3.2.1. Baseline Regression Model

To assess the direct impact of Artificial Intelligence (AI) on both Explicit (EET) and Implicit (IET) Energy Transitions, this study specifies the following two-way fixed effects panel regressions:

$$EE{T}_{i,t}=\alpha +{\beta }_{1}A{I}_{i,t}+{\gamma }^{\prime }{X}_{i,t}+{\mu }_i+{\lambda }_t+{\epsilon }_{i,t}$$

(1)

$$IE{T}_{i,t}=\alpha +{\beta }_{2}A{I}_{i,t}+{\gamma }^{\prime }{X}_{i,t}+{\mu }_i+{\lambda }_t+{\epsilon }_{i,t}$$

(2)

where $EE{T}_{i,t} \mathrm{and} IE{T}_{i,t}$ denote Explicit and Implicit Energy Transition in country i at time t. $A{I}_{i,t}$ measures the intensity of Artificial Intelligence applications. $X_{i,t}$ is a vector of control variables, including economic growth (lnEG), financial development (FD), trade openness (TR), and education (EDU). ${\mu }_i \mathrm{and} {\lambda }_t$ represent country and year fixed effects, respectively, while ${\epsilon }_{i,t}$ is the idiosyncratic error term. Driscoll–Kraay (DK) standard errors are employed to address cross-sectional dependence and serial correlation.

3.2.2. Mediation Effect Model

Since Implicit Energy Transition (IET) may act as a channel through which AI influences explicit transition (EET), we employ a mediation effect framework:

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

(3)

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

(4)

Here, IET_i,t functions as the mediator, capturing AI’s indirect contribution to explicit transition via efficiency gains, smart grids, and energy optimization.

3.2.3. Moderating Effect Model

To test whether Internet Connectivity (INTC) and Knowledge Management (KM) moderate the AI–transition nexus, interaction terms are introduced:

$$EE{T}_{i,t}={\alpha }_0+{\alpha }_{1}A{I}_{i,t}+{\alpha }_{2}INT{C}_{i,t}+{\alpha }_3\left(A{I}_{i,t}\times INT{C}_{i,t}\right)+{\alpha }_4{X}_{i,t}+{\mu }_i+{\lambda }_t+{\epsilon }_{i,t}$$

(5)

$$IE{T}_{i,t}={\beta }_0+{\beta }_{1}A{I}_{i,t}+{\beta }_{2}K{M}_{i,t}+{\beta }_3\left(A{I}_{i,t}\times K{M}_{i,t}\right)+{\beta }_4{X}_{i,t}+{\mu }_i+{\lambda }_t+{\epsilon }_{i,t}$$

(6)

where the coefficients of the interaction terms (α₃, β₃) capture whether the presence of strong digital infrastructure (INTC) or robust knowledge systems (KM) amplifies or mitigates AI’s effect on transition outcomes.

3.2.4. Nonlinear Specification

Finally, to examine whether the effects of AI, INTC, and KM exhibit nonlinear patterns (e.g., inverted-U), we extend the model to include squared terms:

$$EE{T}_{i,t}={\alpha }_0+{\alpha }_{1}A{I}_{i,t}+{\alpha }_{2}A{I}_{i,t}^2+{\alpha }_{3}INT{C}_{i,t}+{\alpha }_{4}INT{C}_{i,t}^2+{\alpha }_5{X}_{i,t}+{\mu }_i+{\lambda }_t+{\epsilon }_{i,t}$$

(7)

$$IE{T}_{i,t}={\beta }_0+{\beta }_{1}A{I}_{i,t}+{\beta }_{2}A{I}_{i,t}^2+{\beta }_{3}K{M}_{i,t}+{\beta }_{4}K{M}_{i,t}^2+{\beta }_5{X}_{i,t}+{\mu }_i+{\lambda }_t+{\epsilon }_{i,t}$$

(8)

The inclusion of quadratic terms allows us to test for diminishing or threshold effects, such as whether internet expansion or knowledge accumulation initially supports but later hinders transition progress, consistent with the “ICT rebound” or “inverted-U innovation” hypotheses.

4. Findings and Discussions

4.1. Descriptive Statistics

Table 6. Descriptive statistics.

Variable	Obs	Mean	Std. Dev.	Min	Max
EET	115	−0.267	0.826	−2.000	0.552
IET	115	−0.383	0.947	−2.000	1.045
AI	115	−0.067	1.013	−1.726	2.000
KM	115	0.081	0.775	−0.985	2.000
INTC	115	0.175	0.889	−1.061	2.000
lnEG	115	8.498	0.758	6.629	9.378
FD	115	77.973	42.944	16.838	181.78
TR	115	44.267	11.714	22.106	68.094
UP	115	1.805	1.167	−0.467	4.198
EDU	115	4.317	1.010	2.038	6.549

presents the descriptive statistics of the variables. The dependent variables, EET and IET, both display negative means (−0.267 and −0.383, respectively), suggesting that on average, the indicators tend to fall below their standardized baselines, though their ranges indicate considerable variation, spanning from –2.000 up to 0.552 for EET and 1.045 for IET. AI, with a mean close to zero (−0.067) and a standard deviation of 1.013, which reflects a relatively balanced variation between its minimum (–1.726) and maximum (2.000) values. Similarly, KM and INTC are centered around small positive means (0.081 and 0.175, respectively), each showing moderate dispersion, indicating heterogeneity in innovation-related capacities. The control variable lnEG averages 8.498, with relatively low variation (σ = 0.758), suggesting stability across the panel. In contrast, FD shows substantial dispersion (mean = 77.973, σ = 42.944), ranging widely from 16.838 to 181.78. TR averages 44.267, with a variation between 22.106 and 68.094, while the UP shows a mean of 1.805 with a wider variability (−0.467 to 4.198), indicating differing levels of urbanization intensity. Finally, EDU displays a mean of 4.317 with moderate variation, ranging from 2.038 to 6.549.

The correlation results in

Table 7. Correlation results.

	EET	IET	INTC	KM	lnEG	FD	TR	EDU	UP
EET	1.000
IET	0.809	1.000
INTC	−0.157	−0.039	1.000
KM	−0.048	0.214	0.272	1.000
lnEG	−0.334	0.127	0.660	0.324	1.000
FD	0.410	0.466	0.250	0.350	0.261	1.000
TR	0.767	0.775	−0.079	−0.100	−0.073	0.210	1.000
EDU	−0.375	−0.426	0.538	−0.342	0.408	0.090	−0.279	1.000
UP	0.318	0.094	−0.419	0.119	−0.541	0.500	0.011	−0.282	1.000

reveal important associations of the EET with other variables. The EET is strongly and positively correlated with the IET (0.809) and TR (0.767), suggesting that implicit transition dynamics and openness to trade relate closely to explicit transition outcomes. It also shows a moderate positive correlation with FD (0.410) and the UP (0.318), highlighting the supportive roles of finance and urbanization in driving transition. Conversely, the EET is negatively correlated with lnEG (−0.334) and EDU (−0.375), implying potential trade-offs where higher growth and education levels are not immediately aligned with explicit transition measures. Correlations with KM (−0.048) and INTC (−0.157) are weak.

4.2. Cross-Section Independence (CD) Test Result

Table 8. Cross-section independence (CD) test.

Variables	CD Test	Mean ρ	Mean Abs(ρ)
EET	5.858 ***	0.39	0.54
IET	7.174 ***	0.41	0.58
INTC	14.814 ***	0.98	0.98
KM	11.878 ***	0.78	0.78
lnEG	13.482 ***	0.89	0.89
FD	6.9490 ***	0.46	0.57
TR	16.694 ***	0.25	0.33
EDU	8.070 ***	0.53	0.53
UP	11.149 ***	0.41	0.65

Note: *** p < 0.01.

reports the results of the cross-sectional dependence (CD) test, where the null hypothesis assumes cross-sectional independence (i.e., no correlation across units in the panel). For all variables, the CD test statistics are highly significant at the 1% level (***), meaning the null hypothesis is rejected in every case. This indicates strong evidence of cross-sectional dependence, as reflected in relatively high average correlations (mean ρ and mean |ρ|), particularly for INTC (ρ = 0.98), lnEG (ρ = 0.89), and KM (ρ = 0.78), suggesting that shocks in one cross-sectional unit are likely to spill over to others.

4.3. Stationarity Test Result

Table 9. CADF test.

	Level	p-Value	First Difference	p-Value
EET	−1.090	0.939	−2.777 ***	0.010
IET	−1.447	0.765	−2.782 ***	0.009
INTC	−1.334	0.837	−3.233 ***	0.000
KM	−1.393	0.801	−2.488 **	0.047
lnEG	−2.035	0.263	−2.690 **	0.016
FD	−2.118	0.205	−2.620 **	0.024
TR	−0.572	0.997	−2.497 **	0.045
EDU	−2.015	0.278	−2.849 ***	0.006
UP	−1.693	0.562	−3.403 ***	0.000

Note: ** p < 0.05, *** p < 0.01.

presents the CADF unit root test, where the null hypothesis assumes that each variable contains a unit root (non-stationary). At levels, all variables have large p-values (greater than 0.10), so the null cannot be rejected, indicating non-stationarity. However, after the first differencing, the test statistics become highly significant for all variables (p < 0.05 or p < 0.01), leading to the rejection of the null hypothesis. This implies that the EET, IET, INTC, KM, lnEG, FD, TR, EDU, and UP are non-stationary at levels but become stationary after the first differencing, confirming they are integrated in order one, I(1).

4.4. Baseline Regression Results

In the baseline model (see

Table 10. Driscoll–Kraay (DK) standard errors (baseline regression) result.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)
	EET	IET	EET	EET	IET	EET	IET
AI	−0.0446	−0.5511 ***	0.1348 **	−0.0321	−0.4705 ***	−0.0530	−0.2939 **
	(−1.0704)	(−6.7419)	(3.2697)	(−0.6883)	(−5.2374)	(−0.8225)	(−2.4945)
lnEG	−0.2808 ***	−0.6510 ***	−0.0689	−0.1560 *	−0.1038	−0.3082 **	−0.8401 ***
	(−5.1099)	(−7.0324)	(−1.1527)	(−1.6772)	(−0.4677)	(−3.0361)	(−4.5917)
FD	0.0012	0.0018	0.0006	0.0014	0.0047 **	0.0009	0.0036
	(1.3874)	(1.2421)	(0.8537)	(1.1912)	(2.4195)	(0.6442)	(1.4249)
TR	0.0027 *	0.0133 ***	−0.0016	0.0044 **	0.0174 ***	0.0026	0.0022
	(1.8216)	(5.3680)	(−1.0689)	(2.7336)	(5.8014)	(1.1024)	(0.5921)
EDU	0.0105	−0.0713 **	0.0337 *	0.0448 **	0.0174	0.0073	−0.0676 *
	(0.5773)	(−1.9958)	(1.8758)	(2.1166)	(0.3549)	(0.3407)	(−1.9119)
IET			0.3255 ***
			(7.6332)
KM				−0.2189 **	−0.4446 **
				(−2.0504)	(−2.0086)
AI*KM				0.0229	−0.0902
				(0.4463)	(−1.0349)
INTC						0.0067	−0.5206 ***
						(0.0893)	(−4.0746)
AI*INTC						0.0083	0.0393
						(0.3211)	(0.7930)
CONS	1.8558 ***	4.6886 ***	0.3295	0.5582	−0.6517	2.1278 **	6.7418 ***
	(3.7480)	(5.9798)	(0.6333)	(0.6411)	(−0.3131)	(2.1283)	(3.8001)
N	115	115	115	115	115	115	115
R2	0.9928	0.9818	0.9954	0.9931	0.9836	0.9928	0.9856

Note: * p < 0.1, ** p < 0.05, and *** p < 0.01. ( ) denotes t statistics.

), Artificial Intelligence (AI) shows a negative but insignificant effect on the Explicit Energy Transition (EET) (−0.0446). This suggests that while AI adoption is growing in BRICS, it has not yet translated into tangible improvements in the explicit transition, partly because digitalization raises electricity demands and fossil fuel use. Economic growth (lnEG) is strongly negative (−0.2808 ***), confirming that fossil-fuel-driven expansion in BRICS slows the pace of the energy transition. Trade openness (TR) is positive and marginally significant (0.0027 *), implying that integration into global markets encourages cleaner technologies through spillovers. Similar findings were reported by [7], who showed that trade fosters renewable energy deployment in emerging economies, while growth tends to hinder it.

When the Implicit Energy Transition (IET) is examined, AI exerts a strong and negative impact (−0.5511 ***), indicating that the indirect effects of digitalization—such as the higher energy demand from data centers—currently outweigh sustainability benefits. lnEG again has a significant negative effect (−0.6510 ***), showing that BRICS’ industrial growth paths are carbon-intensive. Trade openness (0.0133 ***) is highly positive, supporting the role of globalization in diffusing low-carbon technologies. Conversely, education (EDU) shows a negative effect (−0.0713 **), perhaps reflecting that education systems in BRICS remain oriented toward traditional industrial skills rather than green competencies. These results are consistent with [3], who highlighted the environmental drawbacks of rapid economic growth in emerging markets, and [31], who showed trade’s positive influence on clean energy adoption.

Introducing the IET as a mediator changes the relationship between AI and the EET: AI becomes positive and significant (0.1348 **), while the IET itself is strongly positive (0.3255 ***). This implies that although AI directly hinders energy transition, when its role in shaping indirect transition mechanisms is recognized—such as optimizing energy efficiency, smart grids, or predictive maintenance—it contributes positively to explicit energy outcomes. Education (0.0337 *) also turns significant, suggesting that human capital matters more when aligned with AI and indirect transition. This finding aligns with [34], who found that digital innovations support renewable adoption when coupled with efficiency pathways.

Here, Knowledge Management (KM) is introduced as a moderator. KM itself is negative and significant (−0.2189 **), suggesting that current knowledge systems in BRICS are fragmented and not yet fully supporting transition policies. The interaction between AI and KM is insignificant, showing that AI deployment has not been effectively integrated into KM practices. lnEG remains negative, while trade openness is positive and significant (0.0044 **). Education also exerts a positive effect (0.0448 **), indicating that knowledge and skills development can offset some structural barriers. Reference [10] similarly found that weak knowledge systems and institutional frameworks reduce the effectiveness of digital tools in supporting sustainability in emerging economies.

When focusing on the IET, KM remains negative and significant (−0.4446 **), highlighting systemic inefficiencies in knowledge absorption and innovation ecosystems across BRICS. AI continues to exert a negative and significant influence (−0.4705 ***), while the AIKM interaction remains insignificant. Trade openness remains strongly positive (0.0174 **), again pointing to global integration as a channel for indirect transition. This pattern resonates with [35], who emphasized that trade integration fosters clean technologies in emerging economies, while weak innovation systems limit knowledge diffusion.

Internet Connectivity (INTC) is then tested as a moderator. INTC is insignificant (0.0067), implying that while the digital penetration is expanding, it has not directly influenced the Explicit Energy Transition in BRICS. AI also remains insignificant, while lnEG is negative and significant (−0.3082 **). Trade and education are insignificant in this model, reflecting weaker complementarity effects. This suggests that the internet growth in BRICS has mainly driven the consumer demand rather than aligning with sustainability objectives. Reference [36] found similar results, showing that ICT penetration often increases the energy demand in developing economies unless tied to renewable integration.

Finally, when the IET is the dependent variable, INTC has a strong and negative effect (−0.5206 ***), suggesting that rising digital connectivity in BRICS indirectly drives higher energy consumption, consistent with increased data traffic, cloud services, and digital infrastructures. AI also remains negative and significant (−0.2939 **), while lnEG exerts the strongest negative effect (−0.8401 ***). Education is negative and marginally significant (−0.0676 *), indicating a misalignment between higher education outcomes and sustainability needs. These findings align with [37], who argued that digitalization in BRICS often increases carbon emissions in the absence of strong clean energy integration.

4.5. Robustness Check

Table 11. Driscoll–Kraay (DK) standard errors (robustness check) results.

	(1)	(2)	(3)	(4)	(5)	(6)
	Replace and Add Control Variables			AI and Controls Lag One Period
	EET	IET	EET	EET	IET	EET
AI	−0.0480	−0.5298 ***	0.1323 **
	(−1.1202)	(−6.1748)	(3.0806)
lnEG	−0.2625 ***	−0.7654 ***	−0.0019
	(−4.0816)	(−6.8325)	(−0.0259)
FD	0.0015	0.0003	0.0014 *
	(1.4806)	(0.1681)	(1.7668)
TR	0.0031 *	0.0109 ***	−0.0006
	(1.8567)	(3.6315)	(−0.3901)
EDU	0.0170	−0.1121 **	0.0552 **
	(0.7686)	(−2.6712)	(2.8951)
UP	0.0170	−0.1061	0.0531 *
	(0.5362)	(−1.6509)	(1.8248)
IET			0.3404 ***
			(8.0502)
L.AI				−0.0228	−0.4415 ***	−0.0228
				(−0.5638)	(−4.6853)	(−0.5638)
L.lnEG				−0.2975 ***	−0.6279 ***	−0.2975 ***
				(−4.2049)	(−5.3973)	(−4.2049)
L.UP				0.0082	0.0404	0.0082
				(0.2835)	(0.7859)	(0.2835)
L.FD				0.0017 *	0.0042 **	0.0017 *
				(1.8770)	(2.5440)	(1.8770)
L.TR				0.0026 *	0.0169 ***	0.0026 *
				(1.7274)	(6.7422)	(1.7274)
C	1.6051 **	6.2574 ***	−0.5249	1.9937 **	3.7482 **	1.9937 **
	(2.3611)	(5.1112)	(−0.7164)	(3.0113)	(3.4128)	(3.0113)
N	115	115	115	110	110	110
R2	0.99	0.98	0.97	0.95	0.98	0.94

Note: * p < 0.1, ** p < 0.05, and *** p < 0.01. ( ) denotes t statistics.

presents the robustness check results. In Model (1), the EET with extended controls, AI remains negative (−0.0480) but insignificant, suggesting that digitalization alone does not guarantee improvements in explicit transition outcomes. Economic growth (lnEG) has a strong and negative effect (−0.2625 ***), consistent with fossil-fuel-led growth patterns in BRICS that constrain sustainability. Trade openness (TR) is positive and significant (0.0031 *), reinforcing the role of international integration in stimulating cleaner energy practices. Education (EDU) and the urban population (UP) are positive but insignificant, indicating that while human capital and urbanization may support transition, their effects are not yet systematic.

For Model (2), the IET with extended controls, AI again exerts a strong negative impact (−0.5298 ***), confirming that indirect pathways of digitalization currently increase the energy demand and emissions in BRICS. lnEG remains significantly negative (−0.7654 ***), pointing to the continued carbon-intensive nature of growth. TR (0.0109 ***) is positive and significant, showing that trade liberalization is beneficial for indirect transition mechanisms. However, EDU is negative (−0.1121 **) and significant, implying that education systems may not yet be geared toward sustainability-driven innovation. The UP is negative but insignificant, suggesting mixed urbanization effects.

For Model (3), the mediation of the IET in the EET, AI turns positive and significant (0.1323 **), while the IET itself strongly enhances the EET (0.3404 ***). This demonstrates that AI supports the explicit transition only when its indirect role through efficiency gains and technological improvements is accounted for. EDU is positive and significant (0.0552 **), suggesting that human capital complements AI in supporting the EET when mediated through indirect transition pathways. The UP is also positive (0.0531 *) and significant, indicating that urbanization fosters the EET in this mediated framework, likely because denser urban environments create a demand for smarter and cleaner energy solutions [38].

Models (4) and (5) considered the lagged effects with the EET and IET. Lagged AI (L.AI) is insignificant for the EET (−0.0228) but negative and strongly significant for the IET (−0.4415 ***), implying that the rebound effects of digitalization on the indirect transition persist over time. Lagged lnEG is negative and significant in both models, confirming that fossil fuel growth continues to hinder transition even with time lags. Lagged FD is positive and significant in both cases, indicating that financial systems gradually support transition through funding for renewable projects and infrastructure. Lagged TR is also positive and significant and is particularly strong for the IET (0.0169 ***), showing that global integration exerts long-run benefits for the transition. Regarding Model (6), the EET with lagged controls, AI remains insignificant (−0.0228), suggesting that the direct role of digitalization in the explicit transition remains limited even with time adjustments. Lagged lnEG is strongly negative (−0.2975 ***), showing that the negative effect of the fossil-fuel-led growth is robust over time. Lagged FD and TR remain positive and significant, confirming that finance and trade continue to provide structural support for transition. The EDU and UP lagged effects are insignificant, reinforcing the notion that structural reforms are needed to align education and urbanization with sustainability

4.6. Nonlinear Effect of Knowledge Management and Internet Connectivity

This study also employed Lewbel IV-2SLS to check the nonlinear effect of both Knowledge Management and Internet Connectivity (see

Table 12. Nonlinear effect of Knowledge Management and Internet Connectivity (Lewbel IV-2SLS) results.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)
	EET	IET	EET	IET	EET	IET	EET	IET	EET	IET
AI	−0.381 ***	−0.232 ***	−0.799 ***	−0.191	−0.273 ***	−0.113	−0.947 ***	−0.565 ***	−0.206 ***	−0.082
	(0.073)	(0.080)	(0.191)	(0.185)	(0.079)	(0.082)	(0.176)	(0.079)	(0.079)	(0.073)
lnEG	−0.445 ***	0.298 ***	−0.851 ***	0.190	−0.764 ***	0.025	−0.934 ***		−0.977 ***
	(0.062)	(0.066)	(0.142)	(0.151)	(0.083)	(0.097)	(0.150)		(0.101)
FD	0.009 ***	0.009 ***	0.009 ***	0.007 ***	0.009 ***	0.008 ***	0.008 ***	0.012 ***	0.009 ***	0.008 ***
	(0.001)	(0.001)	(0.001)	(0.001)	(0.001)	(0.001)	(0.002)	(0.002)	(0.001)	(0.001)
TR	0.029 ***	0.044 ***	0.026 ***	0.052 ***	0.034 ***	0.049 ***	0.022 ***	0.041 ***	0.037 ***	0.051 ***
	(0.005)	(0.005)	(0.005)	(0.005)	(0.004)	(0.005)	(0.006)	(0.004)	(0.004)	(0.004)
EDU	−0.409 ***	−0.493 ***	−0.260 ***	−0.278 ***	−0.315 ***	−0.390 ***	−0.292 ***	−0.341 ***	−0.325 ***	−0.408 ***
	(0.062)	(0.070)	(0.066)	(0.081)	(0.066)	(0.073)	(0.065)	(0.062)	(0.059)	(0.061)
	(0.243)	(0.242)	(0.256)	(0.270)	(0.201)	(0.221)	(0.258)	(0.259)	(0.217)	(0.232)
KM			0.715 ***	0.257			0.800 ***	0.775 ***
			(0.219)	(0.235)			(0.242)	(0.090)
INTC					0.637 ***	0.555 ***			1.202 ***	0.744 ***
					(0.116)	(0.132)			(0.212)	(0.112)
KM2							0.098	−0.266 ***
							(0.077)	(0.076)
INTC2									−0.251 ***	−0.138 **
									(0.077)	(0.068)
C	3.423 ***	−3.556 ***	6.438 ***	−3.996 ***	4.597 ***	−2.729 ***	7.538 ***	−1.684 ***	6.156 ***	−2.427 ***
	(0.859)	(0.919)	(1.460)	(1.346)	(0.737)	(0.844)	(1.549)	(0.553)	(0.718)	(0.506)
N	115	115	115	115	115	115	115	115	115	115
R2	0.85	0.86	0.89	0.88	0.89	0.88	0.89	0.89	0.90	0.89

Note: ** p < 0.05, and *** p < 0.01. ( ) denotes t statistics.

). In Model (1), the EET baseline, AI shows a strong negative effect (−0.381 ***), suggesting that in BRICS, digitalization increases the energy demand without directly advancing explicit transition. lnEG also has a negative effect (−0.445 ***), reflecting a fossil fuel dependence in growth processes. By contrast, FD (0.009 ***), TR (0.029 ***), and EDU (−0.409 ***) highlight the structural mix: finance and trade promote the EET, while education appears misaligned with transition needs. These results agree with [39], who found that growth worsens emissions in BRICS, while trade supports renewable adoption.

In Model (2), the IET baseline, AI remains negative and significant (−0.232 ***) but is weaker than in the EET, implying that AI adoption indirectly undermines transition through a higher energy intensity. lnEG turns positive (0.298 ***), suggesting that economic growth fosters certain implicit transition processes, possibly through efficiency gains or infrastructure modernization. FD and TR are both positive and highly significant, underscoring their supportive role. However, EDU remains negative (−0.493 ***), showing that educational systems in BRICS have not yet adapted to sustainability-driven innovation. This pattern is consistent with [40], who noted that structural policies matter more than growth for sustainable transition.

In Model (3), the EET with KM, AI’s effect becomes even more negative (−0.799 ***), while KM shows a strong positive effect (0.715 ***), meaning effective KM can offset AI’s negative pressure on transition. Ln EG remains negative (−0.851 ***), while FD and TR continue to support transition. EDU remains negative (−0.260 ***). This demonstrates that BRICS can benefit from knowledge flows and innovation management in overcoming digitalization’s rebound effects. Ref. [20] similarly emphasized the role of KM in enhancing green innovation outcomes.

In Model (4), when an IET is the outcome, KM’s effect is positive though insignificant (0.257), indicating limited evidence of KM’s direct impact on indirect transition channels. AI remains negative but less significant (−0.191), suggesting that KM partly moderates AI’s harmful effects. lnEG turns insignificant, while FD and TR remain positive. EDU is again negative (−0.278 ***). This suggests that while KM may moderate AI’s influence, its role in indirect transition is weaker. Ref. [16] argued that weak knowledge systems in emerging economies limit sustainability gains.

In Model (5), introducing Internet Connectivity (INTC), AI is negative and significant (−0.273 ***), while INTC shows a strong positive effect (0.637 ***), suggesting that digital infrastructure boosts explicit transition by improving access to technologies and efficient systems. lnEG remains negative (−0.764 ***), while FD and TR are supportive. EDU is negative (−0.315 ***). This implies that internet expansion can directly foster transition if accompanied by supportive institutions. Ref. [36] found that ICT enhances sustainability when aligned with clean energy strategies.

In Model (6), the IET with INTC, INTC remains strongly positive (0.555 ***), while AI turns insignificant (−0.113), showing that internet access helps reduce AI’s negative effects on indirect transition. lnEG becomes insignificant, while FD and TR remain strong. EDU is negative (−0.390 ***). This implies that INTC enhances the indirect pathways of transition, such as information flows and efficiency. Similar results were reported by [4], who highlighted the importance of digital connectivity in promoting environmental efficiency.

Pertaining to Model (7), the EET with KM nonlinearity, AI remains strongly negative (−0.947 ***) and KM is strongly positive (0.800 ***), but KM² is insignificant (0.098). This suggests a positive linear effect of KM without diminishing or threshold effects on the EET. lnEG remains negative (−0.934 ***), while FD and TR remain supportive. EDU is still negative (−0.292 ***). This indicates that in BRICS, more KM consistently supports explicit transition, with no evidence of diminishing returns. Ref. [41] confirmed the linear benefits of innovation and KM for renewable adoption.

Moreover, looking at Model (8), the IET with KM nonlinearity, KM remains positive and significant (0.775 ***), but KM² is negative and significant (−0.266 ***), showing an inverted-U relationship: at early stages, KM supports transition, but excessive KM without application efficiency may reduce gains. AI remains negative (−0.565 ***), while FD and TR remain supportive. EDU is negative (−0.341 ***). This suggests that BRICS benefit from KM up to a threshold, after which inefficiencies may appear. Ref. [42] also identified inverted-U effects of innovation on environmental sustainability.

Likewise, in Model (9), the EET with INTC nonlinearity, AI remains negative and significant (−0.206 ***), while INTC is positive (1.202 ***) and INTC² is negative (−0.251 ***), showing an inverted-U effect: internet expansion promotes the EET up to a point, but excessive digital growth increases the energy demand. lnEG remains negative, FD and TR are positive, and EDU is negative. This aligns with [43], who showed that ICT expansion can worsen emissions when infrastructure is carbon-intensive.

Lastly, Model (10), the IET with INTC nonlinearity, AI turns insignificant (−0.082), while INTC is positive (0.744 ***) and INTC² is negative (−0.138 **), confirming the inverted-U relationship for digitalization’s impact on indirect transition. At moderate levels, connectivity fosters information exchange and cleaner practices, but at high levels, the rising energy demand from digital infrastructures outweighs benefits. lnEG is dropped, but FD and TR remain supportive, while EDU is negative. This supports the “ICT rebound hypothesis” in BRICS, consistent with [44]. Figure 1 shows the summary of the results.

In summary, across the 10 models, AI consistently hinders transition unless mediated by KM and INTC. KM shows a linear positive effect on the EET and an inverted-U effect on the IET, while INTC demonstrates inverted-U effects on both EET and IET. Growth consistently constrains the transition, while FD and TR act as robust drivers. Education remains a persistent barrier, reflecting a misalignment of human capital with sustainability goals. Figure 1 presents a summary of the findings.

5. Conclusions and Policy Directions

5.1. Conclusions

This study explores the moderating roles of Internet Connectivity and Knowledge Management in shaping the relationship between Artificial Intelligence and the energy transition in BRICS economies, using annual data from 2000 to 2022. By integrating Driscoll–Kraay (DK) standard errors with Lewbel IV-2SLS estimators, the analysis not only addresses potential endogeneity and cross-sectional dependence but also contributes novel insights into how digitalization and knowledge systems interact to influence the pathways of the Explicit and Implicit Energy Transition. The findings reveal that in BRICS nations, Artificial Intelligence generally exerts a negative effect on both the Explicit and Implicit Energy Transition, largely due to the increased energy demand from digitalization, while economic growth also constrains transition through fossil fuel dependence. In contrast, financial development and trade openness consistently promote both transition channels, underscoring the role of structural integration. Knowledge Management enhances the explicit transition and shows an inverted-U effect on the implicit transition, suggesting efficiency thresholds, while Internet Connectivity similarly displays an inverted-U pattern—boosting transition at moderate levels but reversing at high penetration due to rebound effects. Education, however, remains misaligned with sustainability goals, often exerting a negative influence across models.

5.2. Policy Initiatives

Align AI with Clean Energy and Transition Goals: The results consistently show that Artificial Intelligence (AI) currently exerts a negative effect on both the Explicit and Implicit Energy Transition in BRICS, largely due to rebound effects such as the rising electricity demand from data centers and digital infrastructure. Policymakers should therefore align AI deployment with clean energy objectives by mandating renewable-powered data centers, incentivizing AI-driven energy efficiency applications, and supporting smart grid integration. Regulatory frameworks should also prioritize AI applications that optimize renewable generation forecasting, demand-side management, and energy storage, ensuring that digitalization supports rather than undermines transition objectives.

Strengthen Knowledge Management and Education Systems: The findings highlight that Knowledge Management (KM) has the potential to support the energy transition, but current systems remain fragmented, while education exerts a negative effect due to its misalignment with sustainability needs. BRICS policymakers should invest in building integrated knowledge platforms that connect universities, industries, and governments to diffuse green technologies. At the same time, education curricula should be reoriented toward green competencies, renewable energy engineering, and sustainability-driven digital innovation. Tailored skill development programs can equip the labor force to adapt to the evolving clean energy economy, reducing the mismatch between human capital formation and transition requirements.

Leverage Internet Connectivity Responsibly: Internet Connectivity (INTC) demonstrates both positive and inverted-U effects, suggesting that moderate levels foster transition, but excessive penetration increases the energy demand. To maximize benefits, BRICS nations should encourage ICT expansion that is directly tied to energy efficiency and clean energy deployment—such as smart cities, the digital monitoring of energy systems, and e-governance platforms for sustainable practices. Policymakers must simultaneously regulate the carbon intensity of ICT infrastructure by promoting renewable-powered networks, green cloud services, and energy-efficient digital devices. This will allow connectivity to enhance the transition while minimizing rebound effects.

Reinforce Trade and Financial Channels for Green Transition: Trade openness (TR) and financial development (FD) consistently show positive effects, underscoring their role as critical enablers of transition. BRICS governments should expand green trade agreements, harmonize environmental standards with global partners, and incentivize clean energy imports to accelerate technology transfer. In parallel, financial markets must be mobilized to channel credit, green bonds, and transition-linked financing toward renewable projects and digital sustainability infrastructure. By integrating trade and finance with explicit decarbonization targets, BRICS nations can scale up investment in clean energy and digital–green synergies while reducing their reliance on fossil-fuel-driven growth.

5.3. Managerial Implications

For managers in BRICS, the evidence that Artificial Intelligence (AI) currently exerts negative effects on the energy transition highlights the need for firms to carefully evaluate how digitalization is deployed. Rather than focusing solely on efficiency gains or cost reductions, managers should prioritize AI applications that directly contribute to sustainability outcomes, such as energy optimization, the predictive maintenance of renewable infrastructure, and low-carbon logistics solutions. Firms must also integrate Knowledge Management (KM) systems into their strategic processes, ensuring that innovations and digital tools are diffused across departments and supply chains. This can help offset rebound effects and foster an organizational alignment with transition goals.

The findings also emphasize that education and Internet Connectivity have mixed effects, with education often misaligned and internet expansion showing inverted-U dynamics. Managers should therefore invest in targeted employee training programs to build green competencies and sustainability-driven digital skills rather than relying on traditional education pipelines. At the same time, digital infrastructure should be adopted strategically, focusing on ICT solutions that are powered by renewable energy and designed to minimize energy intensity. Finally, given the positive role of trade openness and financial development, firms should actively leverage international partnerships, green financing instruments, and cross-border technology exchanges to accelerate their transition strategies while maintaining competitiveness in global markets.

5.4. Limitations and Future Directions

This study, while offering robust insights into the nonlinear effects of Artificial Intelligence, Knowledge Management, and Internet Connectivity on the energy transition in BRICS, is not without limitations. First, the analysis relies on aggregate national-level data, which may obscure sectoral or regional heterogeneities, such as differences between energy-intensive industries and service-based economies. Second, the proxies used for AI, KM, and Internet Connectivity may not fully capture the qualitative aspects of digitalization and innovation practices. Third, the models focus primarily on BRICS, limiting the generalizability to other emerging or developed economies with distinct structural and institutional contexts. Future research could extend the analysis by incorporating micro-level firm data, examining sector-specific pathways, and employing machine learning-based methods to capture nonlinearities more dynamically. Moreover, comparative studies beyond BRICS, or longitudinal analyses incorporating post-2024 data, would help validate and broaden the applicability of the findings.