Abstract

This paper examines the extent to which artificial intelligence (AI) can assist the transition from dependence on fossil fuels to renewable energy resources. The study evaluates both the benefits and environmental costs of AI through the frameworks of Hans-Werner Sinn’s Green Paradox and the Sustainability Paradox. Existing research indicates that AI can support renewable-energy systems through improved energy forecasting, grid integration, predictive maintenance, energy storage and system optimisation, potentially increasing the reliability and economic competitiveness of renewable resources. However, AI uses a lot of electricity, water and computing infrastructure, which can add environmental pressure, especially when data centres run on fossil-fuel electricity. Therefore, AI is not automatically a positive fit for the energy transition. Its value depends on whether the environmental and economic gains it creates are greater than the resources it takes to run. With low-carbon power, efficient use of resources and strong environmental policy, AI can help speed up the shift to renewable energy.

Introduction

For most of modern history, economic growth and fossil fuel consumption have moved together almost in lockstep, with oil, coal and natural gas powering everything from electricity grids to transport networks to heavy industry, and for a long time there was little serious alternative to relying on them. That picture has been changing, gradually but unmistakably, as the cost of solar and wind power has fallen dramatically and countries have come under growing pressure, both environmental and economic, to reduce their dependence on fossil fuels, particularly given that energy production remains one of the largest contributors to global greenhouse gas emissions. There are signs that this shift is genuinely underway: in 2025, renewables overtook coal in global electricity generation for the first time in a century, a milestone that would have seemed unlikely even ten years ago. Even so, the transition is far from complete, since fossil fuels still make up around 86% of the world’s total energy supply, and global energy demand continues to rise faster than clean energy can be added to the system.

Artificial intelligence (AI) has entered this picture in a way that complicates the story further, pulling in both directions at once. On one hand, AI itself is a significant and rapidly growing consumer of electricity, with data centre demand growing by 17% in 2025 alone and demand from AI-focused data centres specifically surging by 50%, vastly outpacing the roughly 3% growth seen in overall global electricity demand that year, much of it still met by fossil fuel generation. On the other hand, AI is also being used to help solve some of the very problems the transition has created, forecasting renewable output and electricity demand, managing increasingly complex power grids and improving efficiency in hard-to-decarbonise industries like steel and cement, where AI applications have reportedly cut energy costs by 3-10 percentage points. The International Energy Agency (IEA) has cautioned against leaning too far in either direction, arguing that both the fear that AI will worsen the climate crisis and the hope that it will single-handedly resolve it are likely overstated, given the considerable uncertainty that remains about how much electricity AI will actually require as it scales.

Historical Context

FOSSIL FUELS

Fossil fuels have been used by humans for thousands of years – to power everything from furnaces to cars to electric lights – dating back to ancient civilisations. Fossil fuels are the fossilised remnants of ancient organisms that, like most of the environment, primarily sustained themselves by consuming hydrogen and carbon atoms. The energy stored inside the hydrocarbon-type molecules is what is used as fuel when burned. When the fuel is burned, it releases carbon into the atmosphere, effectively coming to a standstill in the Earth’s atmosphere. This consequently traps the Sun’s heat from being reflected off the Earth, resulting in a global rise in temperature.

INDUSTRIAL REVOLUTIONS AND WORLD WARS

In 1879, James Watt improved a Thomas Newcomen engine, creating an efficient, working steam engine that would kickstart the Industrial Revolution, the beginning of a constant human progression. During the Industrial Revolution, coal took over the global industry. It powered steamboats, factories and steam-powered machines. People’s main form of travel became steam-powered trains and steamboats. Coal replaced wood and charcoal as a heat source. This caused a boom in global economies, which encouraged continued use. Wars further fuelled this, as factories used coal to make weapons. Through these events, coal and oil became some of the most-used resources across the globe and the main sources of energy. Production using fossil fuels increased during the World Wars, with the demand for weapons skyrocketing. Petroleum was in high demand as it was needed to power tanks, fighter planes and warships. Coal was widely used across factories to produce ammunition, military vehicles and other wartime items. Global carbon emissions spiked during and after the wars. The coal and oil industry boomed during this period and stayed high for decades to come.

OIL VS. COAL

In 1885, German engineer Carl Benz invented the first gasoline-powered automobile, forever changing the oil industry. After Henry Ford released his Model T automobile, oil surpassed coal in the United States due to gasoline demand. Oil also overtook coal as the most widespread form of fuel during the World Wars. Later, with oil in such high demand, many large American oil companies were created from the legal breakup of the super-company Standard Oil Company. Many of these companies have since evolved into the United States’ biggest oil majors, transforming oil into one of the most used and coveted global resources.

Modern Usage and Artificial Intelligence

Today, fossil fuels are an integral component of how the modern world functions, powering everything from electricity to factories to transportation. Coal is primarily burned to provide electricity, although the use of coal is rapidly decreasing as it is being overtaken by cheaper alternatives. One such alternative is natural gas, which is used to generate electricity and for heating. Yet, oil remains the world’s most produced and most consumed form of energy. From fuel oil to liquefied petroleum gas to gasoline, it is the most used fossil fuel in the modern day. Its primary use is to power cars, trains, aeroplanes and more for transportation purposes. Despite the growing concerns over the environmental impact and climate change, fossil fuels still remain a crucial component of how society functions.

Notably, artificial intelligence – defined by NASA (2024) as “an artificial system developed in computer software, physical hardware or other context that solves tasks requiring human-like perception, cognition, planning, learning, communication or physical action” – has grown to incorporate nearly every aspect of the modern world. Investors use AI for quicker and more efficient management and research; professionals use AI for calculations and information; and students use AI to study and learn. However, the amount of energy and water required to sustain the data centres and creation of AI is enormous. Models require massive amounts of energy and resources to train and maintain, large amounts of electricity and an enormous number of computational resources. According to Penn State (2026), “data centres consumed 4.4% of US electricity” and is projected to hit near 13% in 2028. Such a large portion of energy poured into AI would put a strain on the national power grid, encouraging more use of fossil fuels to compensate.

Paradoxically, artificial intelligence is also helping reduce its environmental imprint. According to the CFA Institute (2026), AI can help through maintenance, forecasting and optimisation of storage and resources. Its impact on productivity as well as efficiency is helping companies worldwide produce goods and services faster and for a fraction of the electricity and labour cost. At the moment, artificial intelligence is not only hindering the environment but assisting in the fight against the damage.

Institutional Framework

The literature search for this paper aimed to find studies on the extent of which AI assists the transition from the dependence on fossil fuels to the dependence on renewable resources. We approached this question by first asking: how is AI being utilised for fossil fuels, and how is AI being utilised for renewable resources/energy? Through this approach, we were able to find the positive and negative effects of AI in both situations. Sources were found using Google Scholar by using a combination of keywords such as: “artificial intelligence”, “energy consumption”, “energy storage”, “rebound effect”, “Green Paradox” and “Sustainability Paradox”. Sources include the World Trade Organization (WTO), International Monetary Fund (IMF), SBS Research Monograph, Partners Universal International Innovation Journal (PUIIJ) and more. We later researched the development of AI in more depth in order to understand how AI is improving the transition. Through the information gathered, we concurred to analyse how AI would assist in the transition from the dependence on fossil fuels to the dependence on renewable resources through the lens of the Green Paradox and the Sustainability Paradox of AI.

The Green Paradox vs. The Sustainability Paradox

According to economist Hans-Werner Sinn, the Green Paradox describes the undesirable effects that environmental policies can have on emissions. Specifically, environmental policies designed to reduce carbon dioxide emissions and protect the environment can influence fossil-fuel suppliers to alter their production and extraction decisions in anticipation of future restrictions. This concept provides an important framework for evaluating the various ways in which artificial intelligence can assist in the transition from fossil fuels to renewable energy sources. As the use of AI continues to increase, concerns have emerged regarding its environmental costs. Although AI can provide benefits to the development and implementation of renewable energy, it also requires large amounts of electricity, and an increasing reliance on AI can contribute to greater energy demands. Therefore, evaluating AI’s role in the energy transition requires a cost-benefit analysis that considers both AI’s potential to accelerate the adoption of renewable energy and the environmental consequences associated with its increasing energy consumption.

One potential solution to this issue is the inclusion of gradual and predictable environmental policies. These policies encourage a transition toward renewable energy while reducing the incentive for fossil-fuel suppliers to increase production. If suppliers anticipate that fossil fuels will become less profitable because of future environmental regulations, they may respond by increasing extraction and selling their resources before those policies take effect. However, policies that gradually reduce fossil-fuel demand while simultaneously supporting renewable-energy development could limit this response. As the reduction in fossil-fuel demand would occur gradually, suppliers would have less incentive to rapidly increase extraction in response to declining future profitability. The usage of AI for fossil fuel suppliers would be difficult to implement because it would lead to a negative cost-benefit situation. The cost to produce the energy from fossil fuels would be large compared to the profit from the energy. AI could support this transition by improving the efficiency of renewable-energy systems, forecasting energy demand and helping manage electricity grids. By combining environmental policies with AI-driven improvements in renewable energy, governments may be able to reduce fossil-fuel dependence while limiting the unintended effects described by the Green Paradox. 

Similar to the Green Paradox, the Sustainability Paradox describes initiatives that unintentionally harm the environment or social inequalities, ultimately failing their sustainability goals. While the Green Paradox focuses on the unintended effects that environmental policies can have on fossil-fuel suppliers, the Sustainability Paradox provides another perspective by examining the potential conflict between economic development, technological advancement and environmental sustainability.

The Green Paradox provides an important framework for evaluating the environmental consequences of artificial intelligence because both concepts demonstrate how technological or environmental solutions can produce unintended consequences. Hans-Werner Sinn’s Green Paradox explains that environmental policies intended to reduce future fossil-fuel demand can encourage resource owners to accelerate extraction before those policies become more restrictive, potentially increasing emissions in the present (Sinn, 2012). Similarly, Chouksey et al. identify a sustainability paradox surrounding AI, stating that although AI can support environmental protection through renewable-energy management, climate prediction and other applications, its rapid expansion also creates substantial environmental pressures through energy consumption, water demand and electronic waste (Chouksey et al., 2026). The two concepts therefore relate to each other because both demonstrate that an intended environmental improvement does not necessarily produce a purely positive environmental outcome. Based on the research question, AI may accelerate the transition away from fossil fuels by improving renewable-energy systems, but the increasing energy requirements of AI could also increase demand for electricity generated from fossil fuels. Therefore, AI should not be viewed as inherently beneficial or harmful to the energy transition; instead, its overall impact depends on whether the environmental benefits created by AI outweigh the additional resources required to operate it.

AI AS A TOOL FOR THE RENEWABLE TRANSITION

AI has the ability to substantially assist the transition from fossil fuels to renewable resources by improving the efficiency, reliability and economic viability of renewable-energy systems. Ukoba et al. explain that AI can be applied across renewable-energy systems through resource assessment, energy forecasting, system monitoring, control strategies and grid integration. These applications can help address some of the limitations of renewable energy, particularly the variability of sources such as solar and wind power (Ukoba et al., 2024). Rane, Choudhary and Rane similarly identify AI and machine learning as tools for improving renewable-energy forecasting, energy production, distribution, efficiency and predictive maintenance (Rane, Choudhary & Rane, 2024). Energy storage provides another significant usage. Zhang and Strbac identify supervised learning, deep learning, reinforcement learning and neural networks as methods that can be used for battery management, energy forecasting, predicting possible maintenance and system optimisation (Zhang & Strbac, 2025). Their review reports examples of AI applications reducing grid disruptions and operational costs while improving energy-storage management. Consequently, AI possibly made renewable energy more reliable and economically competitive, reducing some of the barriers that currently contribute to dependence on fossil fuels.

THE ENVIRONMENTAL COSTS AND INCONSISTENCIES OF AI

However, the benefits of AI integration into renewable-energy systems must be considered alongside the environmental costs associated with AI itself. Chouksey et al. emphasise that the rapid expansion of AI and large-scale data centres increases pressure on energy, water and other resources, creating a scenario which can be described as the Sustainability Paradox (Chouksey et al., 2026). This creates an important inconsistency in AI’s role in the energy transition: AI can reduce energy waste and improve renewable-energy efficiency while simultaneously increasing the amount of energy required for computation and data-centre operations. Therefore, the environmental benefit of an AI application cannot be evaluated independently from the resources needed to operate it. If the additional electricity required by AI is generated using fossil fuels, the expansion of AI could indirectly contribute to continued fossil-fuel demand rather than eliminating it. This creates a potential connection to the Green Paradox because an innovation intended to support environmental improvement could unintentionally create additional demand for fossil-fuel resources. However, this relationship should not be interpreted as meaning that AI automatically causes a Green Paradox; rather, it demonstrates why the source of the energy used by AI is an important factor in determining whether AI accelerates or slows the renewable transition.

THE EQUILIBRIUM BETWEEN AI’S BENEFITS AND RESOURCE CONSUMPTION

Ultimately, the extent to which AI assists the transition to renewable energy depends on achieving an equilibrium between the benefits provided by AI and the amount of resources consumed by AI. The evidence suggests that AI can improve renewable-energy forecasting, grid integration, energy storage, system efficiency and maintenance, potentially allowing renewable resources to become more reliable and economically viable. At the same time, the continual expansion of AI creates additional demands for electricity, water and infrastructure, meaning that the environmental gains of AI cannot be assumed to outweigh its costs. Therefore, this equilibrium can be understood as a comparison between the energy and resources required to operate AI and the energy, emissions and resources saved through AI-enabled optimisation. If AI is increasingly powered by low-carbon electricity, while improving the efficiency and integration of renewable energy, its contribution to the transition could become increasingly significant. On the other hand, if the expansion of AI continues to rely heavily on energy systems dependent on fossil fuels, its environmental benefits may be partially offset by the additional demand it creates. Hence, AI is best to be understood not as a complete solution but a tool used to assist the transition from the dependence of fossil fuels to the dependence of renewable resources based on the advantages and disadvantages.

Discussion

This review concludes that artificial intelligence cannot be assigned a straightforward positive or negative role in the shift from fossil fuels to renewable energy. Though the Green Paradox and the Sustainability Paradox were formulated to account for distinct phenomena, they share a common root: measures designed to yield environmental benefits can produce side effects that undercut those benefits. Sinn’s framework concentrated on how suppliers respond when regulation is anticipated, whereas Chouksey et al.’s Sustainability Paradox examines AI’s own resource consumption, but both expose the same fundamental issue: a solution operating within an economic system is shaped by that system’s competing pressures. AI’s overall impact on the transition is determined less by its technical capabilities than by the circumstances surrounding its use.

The applications surveyed here – forecasting, grid management, predictive maintenance and battery optimisation – all show genuine promise in enhancing the reliability and competitiveness of renewable systems relative to fossil fuels. This view is reinforced by Ukoba et al. and Rane et al., while Zhang and Strbac’s work on storage management indicates that AI can lower the operational costs that have long made renewables less appealing to grid operators. Variability has consistently been the most compelling objection to solar and wind, and AI-based forecasting directly tackles that weakness. Yet the picture becomes more complicated when accounting for AI’s own energy demands, a factor earlier scholarship could not fully foresee. The 2025 figures reveal a 17% increase in overall data centre electricity consumption and a 50% jump specifically among AI-focused facilities, while global electricity demand grew by only 3%. This indicates that AI’s resource use is neither small nor steady; it is expanding more rapidly than the energy system is decarbonising. That trajectory matters because it determines which side of the balance AI occupies at any given moment. Should AI-driven improvements in renewable efficiency be overtaken by the growth of AI’s electricity consumption, and should that consumption rely heavily on fossil generation, then AI would reproduce the very dynamic described by the Green Paradox: a technology intended to support decarbonisation would instead perpetuate demand for the fuels it was designed to displace.

The IEA’s cautionary stance is relevant here. A more nuanced interpretation of the evidence, rather than labelling AI as either a climate solution or a climate burden, is that AI amplifies whatever energy system surrounds it. In a grid that is already largely low-carbon, AI’s computational needs impose a smaller environmental penalty, and its optimisation capabilities can reinforce the grid’s existing advantages. In a grid still dependent on fossil fuels, however, AI’s demand adds directly to that dependency, no matter how much it improves renewable forecasting elsewhere in the system. Consequently, the question of whether AI helps or harms the transition has no fixed answer. It shifts according to the energy mix that powers AI itself, exactly the equilibrium condition identified earlier in this paper.

A further concern involves a mismatch in scale. The energy cost reductions AI achieves in sectors such as steel and cement, reported at 3-10 percentage points, are real but modest. By contrast, the growth in AI’s electricity demand is not modest; it compounds annually. Unless efficiency gains from AI applications expand at a comparable pace, or unless the electricity powering AI systems is decarbonised far more quickly than current projections suggest, there is a genuine danger that AI’s footprint will exceed its contributions. This does not imply that AI is currently net harmful to the transition, given that renewables surpassed coal globally in 2025, but it does mean the present trajectory cannot be assumed to remain favourable without deliberate action.

These findings collectively indicate that AI’s role in the transition is neither fixed nor intrinsic to the technology. It depends on policy decisions, especially regarding how data centres are powered, and on whether governments link AI expansion to the kind of gradual, predictable environmental regulation discussed earlier in connection with the Green Paradox. AI has shown real ability to accelerate certain parts of the transition, yet that ability does not ensure a net positive outcome unless the resource costs of AI are actively managed rather than treated as an afterthought. This review draws primarily on secondary sources, several of which employ differing methodologies and timeframes to estimate AI’s energy consumption, complicating direct comparisons. The work of Chouksey et al. and the IEA data cited here have appeared only recently, so long-term empirical evidence on whether AI’s efficiency gains outstrip its own electricity growth remains scarce. Much of the research on AI applications in renewable energy also derives from pilot studies or industry-reported figures rather than independent verification, meaning that reported benefits, such as the 3-10 percentage point cost reductions in steel and cement, may not apply uniformly across sectors or regions.

Conclusion

Artificial intelligence has a crucial role to play in shifting our reliance from fossil fuels to renewable energy sources, though this transition presents substantial trade-offs. This review highlights how AI can enhance renewable energy systems by refining energy forecasting, optimising grid management, enabling predictive maintenance, improving storage solutions and boosting overall system efficiency. These advancements can tackle some of the key challenges of renewable energy, especially the unpredictability of sources like wind and solar, making them more dependable and cost-effective. However, the rapid growth of AI also brings about significant demands for electricity, water and infrastructure. If this energy comes from fossil fuels, AI could inadvertently reinforce the very dependence it aims to alleviate. 

The Green Paradox and Sustainability Paradox illustrate why we need to consider these conflicting effects together. Both concepts reveal that actions meant to benefit the environment can sometimes lead to unintended consequences that undermine those benefits. Therefore, we should not see AI as a magic solution for climate change or the energy transition. Instead, it is a tool whose success hinges on the energy systems, policies and resources that support it. If AI systems increasingly draw power from renewable and low-carbon energy sources while simultaneously enhancing the efficiency and reliability of renewable systems, we could see a significant boost in the shift away from fossil fuels. In contrast, if AI’s energy consumption keeps rising while still relying on fossil fuel generation, the environmental advantages it offers might be diminished or even negated.

Ultimately, the extent to which AI assists the transition depends on achieving an equilibrium between the resources consumed by AI and the energy and emissions saved through its applications. Current research provides evidence of meaningful opportunities, but long-term evidence remains limited, particularly regarding whether AI’s efficiency gains will consistently outweigh its growing energy requirements. Therefore, the future role of AI in the energy transition will depend not only on technological advancement but also on deliberate environmental policies, continued development of renewable energy and the decarbonisation of the electricity used to power AI. AI cannot independently replace fossil fuels; however, when integrated into a cleaner energy system and supported by effective policy, it can serve as a powerful tool for accelerating the transition toward renewable resources.

British Geological Survey (2022) The greenhouse effect. British Geological Survey [online]. <https://www.bgs.ac.uk/discovering-geology/climate-change/how-does-the-greenhouse-effect-work/>

CFA Institute (2026) Will AI help or hinder the energy transition?, CFA Institute [online]. <https://www.cfainstitute.org/insights/articles/ai-energy-transition-risks-and-opportunities>

Chouksey, A., Rajan, A.K., Gurjar, V., Tiwari, R. and Mishra, P.K. (2026). The green paradox: The climate, environmental, and sustainability implications of artificial intelligence. Global Environmental Change Advances, 6, 100029.

Copeland, B.J. (2024) History of artificial intelligence. Britannica [online]. <https://www.britannica.com/science/history-of-artificial-intelligence>

Environmental and Energy Study Institute (2021) Fossil fuels. Environmental and Energy Study Institute [online]. <https://www.eesi.org/topics/fossil-fuels/description>

Frimpong, V. (2025). The sustainability paradox of artificial intelligence: How AI both saves and challenges resource management efforts. Artificial Intelligence and Sustainability: Innovations in Business and Managerial Practices, pp.60–79.

Fulghum, N., Suarez, W., Altieri, K. and Rangelova, K. (2026) Global Electricity Review 2026. Ember [online]. <https://ember-energy.org/latest-insights/global-electricity-review-2026/>

IEA (2026) Key Questions on Energy and AI (Paris: International Energy Agency).

Kandemir, M. (2025) Why AI uses so much energy—And what we can do about it. Institute of Energy and the Environment [online]. <https://iee.psu.edu/news/blog/why-ai-uses-so-much-energy-and-what-we-can-do-about-it>

Kingsford, P.W. (2018) James Watt. Britannica [online]. <https://www.britannica.com/biography/James-Watt>

Kool, T. (2020) The complete history of fossil fuels. Oil Price [online]. <https://oilprice.com/Energy/Energy-General/The-Complete-History-Of-Fossil-Fuels.html>

Lyman, R. (2026) The 2026 Statistical Review of World Energy. Friends of Science [online]. <https://blog.friendsofscience.org/2026/07/02/the-2026-statistical-review-of-world-energy/>

May, K. (2024) What is artificial intelligence? NASA [online]. <https://www.nasa.gov/what-is-artificial-intelligence/>

Rane, N., Choudhary, S. and Rane, J. (2024). Artificial Intelligence and Machine Learning in Renewable and Sustainable Energy Strategies: A Critical Review and Future Perspectives. Partners Universal International Innovation Journal (PUIIJ), 2(3).

Sinn, H.-W. (2012). Green Paradox. Hans-Werner Sinn [online]. <https://www.hanswernersinn.de/en/topics/GreenParadox>

Smithsonian (2021) What are fossil fuels? Smithsonian [online]. <https://ocean.si.edu/conservation/gulf-oil-spill/what-are-fossil-fuels>

Ukoba, K., Olatunji, K.O., Adeoye, E., Jen, T.-C. and Madyira, D.M. (2024). Optimizing renewable energy systems through artificial intelligence: Review and future prospects. Energy & Environment, 35(7), pp.3833–3879.

United Nations (2025) Artificial intelligence: How much energy does AI use? United Nations Western Europe [online]. <https://unric.org/en/artificial-intelligence-how-much-energy-does-ai-use/>

Zhang, T. and Strbac, G. (2025). Artificial intelligence applications for energy storage: A comprehensive review. Energies, 18(17), 4718.