Artificial Intelligence (AI) on Energy Efficiency: A strategic imperative

The Energy Trilemma—balancing Energy Security, Environmental Sustainability, and Energy Equity—is a key challenge for modern society. AI offers transformative potential in optimizing energy efficiency, especially in oil and gas production, which remains, so far, the dominant energy source despite projected declines. For AI to have a national-scale impact, solutions must be scalable and replicable, targeting high-energy-demand subsystems. AI can also stabilize market volatility, optimize strategic storage, and enhance production trade-offs through bottomhole pressure reservoir control, well connectivity mapping, and risk assessment in unconventional reservoirs to prevent production losses while ensuring resource extraction. By integrating these capabilities, AI becomes a powerful tool for advancing energy efficiency in terms of the Energy Trilemma in a complex and constantly shifting global landscape.

Introduction

The Energy Trilemma: Three interconnected objectives that involve trade-offs, improving one often comes at the expense of another. These objectives are crucial to national interest due to their economic and sustainability implications:

• Energy Security: A balance of import independence, diverse electricity generation, and efficient energy storage. “Reflects nation's capacity to meet current and future energy demand reliability, withstanding and bouncing back swiftly from system shocks with minimal disruptions to supplies” [1].

  
  

• Energy Equity: Equitable access to electricity and fair pricing of essential energy sources, including electricity, gasoline, and diesel. “Assesses a country’s ability to provide universal access to affordable, fairly priced and abundant energy for domestic and commercial use” [1].

  
  

• Environmental Sustainability: A combination of energy efficiency (energy per GDP), low-carbon electricity generation, and reduced per capita CO₂ emissions. “Represents the transition of a country’s energy system towards mitigating and avoiding potential environmental harm and climate change impact” [1].

  
  

This article also explores the critical role of AI in energy and how, when applied responsibly, AI can contribute to addressing the Energy Trilemma. Below is the estimated "balance " of the Energy Trilemma for the U.S., which ranks 10 among the countries considered.

Figure 1: Energy Trilemma for the U.S. and historical performance [4]

Below are the key factors influencing each component and their respective directional variations. For example, regarding Energy Security, import independence has increased, likely due to rising oil production. However, energy storage has declined, possibly because no new reserves have been added, the slowdown in oil production decline is insufficient, or the risk of breakeven deterioration for existing and new oil projects remains high due to geopolitical pressures on oil prices, such as those from OPEC (i).

Figure 2: Factors per trilemma Component [4]

The following scatter plots illustrate the relationships between the three factors of the Energy Trilemma. The color convention from figures 1 and 2 is consistently applied to the axes for each trilemma factor. Additionally, the specific combination for each case in the U.S. is highlighted in purple.

Figure 3: Scatterplot for each combination of Energy Trilemma objectives (ii)

The ranking with the maximum difference between the current score and its maximum possible value of 100 is Energy Security. We will see below that Energy Security is an objective of special interest within the Energy Trilemma because of its implications for economic sustainability.

Efficiency Implies Addressing the Energy Trilemma

The term efficiency means doing more and less at the same time. Or strictly, according to Pareto-Koopmans definition [2], is doing more for a given level of entries, i.e., efficiency oriented to outputs, or doing the same with fewer entries, i.e., efficiency oriented to inputs. For simplified purposes, we will use only the output-oriented schema.

Thus, an efficient operation means that there is no other combination that produces a higher output with the given input levels. For example, consider Energy Security as the output, while Energy Equity and Environmental Sustainability as the inputs. If there is no other possible combination where we can increase Energy Security with the same level of Energy Equity and Environment Sustainability, then we declare that our current combination is efficient. Those combinations that meet this condition form what is known as the efficiency frontier.

The next figure presents the structure for an output-oriented schema, assuming the other two factors are entries. For example, graph (a) presents Energy Security as the output, given a level of entries of Energy Equity and Environment Sustainability. The same scheme is applied for Equity Energy and Environmental Sustainability as outputs, respectively, graphs (b) and (c).

Figure 4: Output efficiency orientation by each trilemma objective combination

By way of illustration, consider in Figure 5 three possible shifts of the current efficiency realization for the case of efficiency oriented to maximize Energy Security. The first case, graph (a), shows a negative efficiency shift, because although Energy Security and Environmental Sustainability increase, Energy Equity deteriorates to a greater extent. Graph (b) presents an innocuous case, since Energy Security and Environmental Sustainability are improved, but the deterioration in Energy Equity cancels the result. Finally, the case of Graph (c) is positive, since it increases Energy Security and Environmental Sustainability, in a proportion greater than the marginal deterioration of Energy Equity. In other words, this last case allows us to argue that we are obtaining higher Energy Security, in proportion to the input levels, Environmental Sustainability, and Energy Equity.

Figure 5: Bad, innocuous, and good efficiency displacements for Energy Trilemma with Security Energy orientation

The challenge lies in identifying technological initiatives that simultaneously advance the three objectives of the Energy Trilemma, ensuring a legitimate contribution to energy efficiency while prioritizing Energy Security. As we will explore later, responsible AI plays a crucial role in this process by structuring initiatives that optimize the efficiency of energy systems.

Energy Security is of special relevance

Among the Energy trilemma objectives, Energy security holds particular importance as it is inherently linked to sustainability. For instance, a lack of energy import independence increases vulnerability to geopolitical pressures, potentially compromising other objectives in the future.

A prime example is the cartelization of OPEC, an organization that excludes the U.S. This group often seeks to drive prices down by intentionally increasing production, given the high elasticity of oil prices to oil tank inventory (due to the high difficulty/cost of storage). Subsequently, oil-field producers with higher breakeven costs are forced to scale back operations. As a result, countries with more easily exploitable oilfields gain greater market control by generating energy dependence and later adjusting prices accordingly.

The Role of AI in Energy Efficiency

Before explaining the role of AI in energy efficiency, it is also good to explain the efficiency of AI (iii). For example, the higher the number of parameters in an AI model, the higher the energy consumption. Therefore, an exaggerated number of parameters that do not produce a significant improvement in the AI model generates an inefficient model.

The same is true for value addition. If AI is used to optimize an energy system, there is a value-adding purpose. On the other hand, if AI is used to convert a real image to a cartoon format, it is surely not the most responsible use, in terms of not generating value addition, but instead tending to waste unnecessary energy.

The other element to consider in producing truly value-added is controlling spurious inferences, lack of generalization, and/or hallucinations of the models. This element is of special care in systems, because generating a false positive can cause a disorientation or loss of efficiency, instead of causing an improvement.

The following figure presents three scenarios. Case (a) refers to the deterioration in efficiency when the AI model produces false positives and/or hallucinations. Case (b) presents the scenario of an AI model with no real purpose, producing a null effect on the operation of the systems. Finally, case (c) represents the effect that responsible AI should have on the systems.

Figure 6: Efficiency displacements equivalences for three scenarios using AI

In summary, Figure 6 for case (c) illustrates a real improvement in overall energy efficiency, contributing to the resolution of the Energy Trilemma.

Why is AI useful for optimizing energy subsystems?

It is no accident that the Department of Energy (DOE) has a policy on AI in energy systems and predicts that it will transform the industry landscape [6] – e.g., on the impact on industrial systems on their energy efficiency, including the energy systems themselves.

The U.S. government agenda underscores the significance of AI development, emphasizing its alignment with national values and its critical role in optimizing systems, particularly energy subsystems, as reaffirmed by the DOE secretary [7]. A key aspect to highlight is the holistic understanding of the energy matrix, recognizing oil and gas as fundamental pillars in meeting energy demands and underscoring the need to enhance their efficiency.

The reason is simple: in a system, the sum of individual parts does not equate to the efficiency of the system as a whole. For example, optimizing an oil and gas well individually differs from optimizing them collectively or simultaneously for the efficient exploitation of the reservoirs. Given that numerous wells can exist in the same field, often interconnected, the complexity of such systems must be carefully considered.

AI plays two essential roles in improving operational efficiency:

1. Mapping the entire system: For example, understanding the interconnections between oil wells and predicting the impact of different actions. This is of special relevance in physical, dynamic, and partially observational systems, such as the majority of energy sub-systems, including oil and gas reservoirs.

   
   

   Likewise, the mapping of the system must consider uncertainty, and an apprentice scheme that allows not only to visualize the best possible information of the now, but also to execute a sufficient design of experimentation to ensure continuous learning and thus sustained optimization actions.  This typically refers to the Reinforcement Learning tradeoff between exploration vs. exploitation, paradoxically inspired by the oil and gas industry. For this reason, it is said that “data analytics will enable a real-time understanding of reservoirs” [8].

2. Determining the optimal course of action: Once the system is fully understood, AI helps define the best sequence of decisions to maximize efficiency using the same level of resources. For example, determine the change of bottomhole pressures to be carried out in each well, and the order in which these actions should be executed, considering resources and constraints. For example, imagine that you know the “chessboard” - or the oil and gas reservoir. We would like to set the actions to be executed within the possibilities, such that the value added is maximized.

   
   

In the case of oil field operations, AI can determine optimal bottom-hole pressures to maximize production and recovery while minimizing energy consumption. This includes reducing water recirculation, whether from natural aquifers or water injections in secondary and tertiary recovery processes. Additionally, AI helps lower the energy required for artificial lifting systems per equivalent barrel of hydrocarbons produced.

In other words, if responsible AI ensures a legitimate improvement in the efficiency of an energy sub-system, this can be translated into a contribution to help solve the Energy Trilemma.

This is what is expressed in the figure below, where an improvement in the energy sub-system of oil extraction, such that internal oil production and its recovery factor are increased without increasing energy consumption level (a win-win optimization), would produce greater Energy Security, while at the same time providing greater Environmental Sustainability and Energy Equity.

Figure 7: Example of responsible IA applied to optimize the efficiency of energy systems and their impact on the energy trilemma

What does it take for AI optimization to have a high impact?

The effectiveness of AI in enhancing national energy efficiency depends on two key factors:

1. Scalability: AI-driven optimization must be replicable across multiple oil and gas fields, ensuring widespread impact. The ability to scale AI solutions efficiently is a major advantage of Responsible AI business cases.

   
   

2. Relevance of the Energy Subsystem: The proportion of energy consumption attributed to the optimized source determines its overall impact. Figure 8 illustrates historical and projected primary energy consumption through 2050.

Figure 8: Energy Production, Consumption, and Shortage by fuel source (iv)

While scalability (point [i]) is essential, its feasibility hinges on a Responsible AI framework, including well-structured algorithms and a business model that enables competitive replication. Therefore, we will focus on the relevance of the energy subsystem (point [ii]).

To achieve a massive impact, AI must optimize high-consumption energy sources. Oil and gas currently account for approximately 78% of total energy consumption, and although this share is expected to decline to 71% by 2050, these resources will remain the dominant energy sources.

In other words, if the goal is to maximize energy efficiency, optimizing oil and gas production should be the top priority, offering the most significant and pragmatic impact on a national scale.

Figure 8 also highlights the historical gap between oil production and demand. While crude oil production has traditionally faced a structural deficit, recent years have seen a surplus, largely driven by unconventional hydrocarbons, which have benefited from high prices. However, this surplus presents two critical considerations:

• Reserves and Well Spacing: Optimizing well interference and communication in unconventional reservoirs is essential to maximize production and reserves, ensuring that no valuable oil is left untapped. AI can play a crucial role by mapping and analyzing subsurface systems, such as fracture heterogeneity, while minimizing the risk of well-to-well communication that could lead to production losses. A tradeoff exists between the risk of inter-well communication, which may reduce output, and the risk of leaving oil unextracted due to wide spacing between wells.

  
  

• Exogenous Factors: External influences like geopolitical and macroeconomic conditions significantly shape production trends. Oil prices, such as WTI, directly impact drilling activity, with higher prices increasing drilling and lower prices discouraging it. According to the EIA, the effect of a price drop is greater than that of a price increase, and production in the Lower 48 states (excluding Alaska and the Federal Offshore Gulf of Mexico) is highly price elastic [10]. This type of asymmetry in price-production elasticity is exploited by organizations such as OPEC and influences the Energy Security factor.

  
  

Given these dynamics, mechanisms that stabilize price volatility and enhance storage management are essential:

• Efficiency Improvements: Increasing efficiency lowers the breakeven point of projects, making them more resilient to price fluctuations.

  
  

• Pressure Management and Strategic Storage: Controlling bottom-hole pressures allows natural reservoir storage to be optimized, helping producers strategically adjust production based on market conditions.

  
  

  • During low-price periods, production should be optimized by prioritizing recovery factors, ensuring long-term sustainability.

    
    

  • During high-price periods, production is maximized immediately, balancing the tradeoff between short-term output and long-term recovery efficiency.

Conclusions

• The energy trilemma remains a critical challenge for modern society, with energy security serving as the foundation for achieving Environmental Sustainability and Energy Equity. Energy efficiency is a key metric in managing this trilemma, requiring an Energy security-oriented approach to balance Environmental Sustainability and Energy Equity.

• Effective measurement is essential; what cannot be measured cannot be controlled or improved. A responsible AI-driven optimization must demonstrate simultaneous efficiency improvements across the three objectives of the Energy Trilemma.

• For AI to have a high impact in optimizing the Energy Trilemma, it must be applied to significant energy subsystems, particularly those with substantial energy consumption. Additionally, scalability is essential, meaning solutions must be both replicable and repeatable with a proper business case or AI deployment scheme.

•  If selecting an energy subsystem for optimization to maximize the national-level impact on the Energy Trilemma, oil and gas extraction would be the highest priority.

• A business case is established for the use of AI algorithms to optimize production and recovery factors, particularly by enhancing water sweep efficiency, mapping well connectivity and interactions, and balancing bottom-hole pressure. This optimization respects operational constraints while preventing excessive energy consumption.

• The previous approach also helps absorb market volatility and can serve as an extension of strategic storage capacity. By better characterizing production responsiveness to market demand or high prices, producers can optimize their output dynamically. Similarly, lower production during price downturns should be strategically managed to avoid deteriorating recovery factor due to unnecessary, unbalanced accelerated extraction, especially in water-drive reservoirs.

• AI can also be leveraged to map the risk of inter-well communication in unconventional reservoirs, enabling large-scale well-fill strategies that ensure no untapped oil is left behind.

• Typically, AI's energy consumption is analyzed, but its responsible use and purpose value generation, such as optimizing energy subsystems, are not always considered. AI should consume energy responsibly, ensuring that the value it generates outweighs the energy it consumes. In other words, energy efficiency principles are universal and should also apply to AI systems.

Footnotes

(i) OPEC: The Organization of the Petroleum Exporting Countries, also known as OPEC, was formed in 1960 by Iraq, Iran, Kuwait, Saudi Arabia, and Venezuela. The organization has since expanded to 13 members. OPEC regularly meets to set oil production targets and coordinate output to help manage global oil prices for the entire group [3].

(ii) Source [5], own calculations and analysis.

(iii) By 2030, the projected energy consumption of data centers, driven by the widespread adoption of AI, is expected to fall within the range of Russia's and India's national energy consumption [11].

(iv) Data Source [10]. Graph and analysis, author elaboration.

References

[1] World Energy Council. (2019). World Energy Council. Worldenergy.org. https://www.worldenergy.org/.

‌[2] Mirdehghan, S. M., & Fukuyama, H. (2016). Pareto–Koopmans efficiency and network DEA. Omega, 61, 78-88.

[3] Hill, Sean, and Owen Comstock. “What Is OPEC+ and How Is It Different from OPEC?” Www.eia.gov, 9 May 2023, www.eia.gov/todayinenergy/detail.php?id=56420.

[4] World Energy Council. “WEC Energy Trilemma Index Tool.” Trilemma.worldenergy.org, 2023, https://trilemma.worldenergy.org/#!/country-profile?country=United%20States&year=2023 

[5] World Energy Council. “WEC Energy Trilemma Index Tool.” Trilemma.worldenergy.org, 2023, https://trilemma.worldenergy.org/#!/energy-index 

[6] U.S. Department of Energy. “Artificial Intelligence for Energy.” Energy.gov, 29 Apr. 2024, www.energy.gov/topics/artificial-intelligence-energy.

[7] “White House Fact Sheet: President Donald J. Trump Is Accelerating America’s Leadership in Artificial Intelligence.” Energy.gov, 12 Feb. 2019, www.energy.gov/articles/white-house-fact-sheet-president-donald-j-trump-accelerating-americas-leadership. Accessed 21 May 2025.

[8] Bromhal, G. (2018). Machine Learning and Data Analytics Overview DOE Upstream Oil and Gas Workshop. https://www.energy.gov/sites/prod/files/2018/02/f49/Machine%20Learning%20and%20Data%20Analytics%20Overview.pdf, page 6.

‌[9] U.S. Energy Information Administration. “U.S. Energy Facts Explained.” Eia.gov, U.S. Energy Information Administration, 15 July 2024, www.eia.gov/energyexplained/us-energy-facts/.

[10] “Short-Term Energy Outlook - U.S. Energy Information Administration (EIA).” Www.eia.gov, www.eia.gov/outlooks/steo/report/perspectives/2023/11-WTIprice/article.php.

[11] “AI Needs More Abundant Power Supplies to Keep Driving Economic Growth.” IMF, 13 May 2025, www.imf.org/en/Blogs/Articles/2025/05/13/ai-needs-more-abundant-power-supplies-to-keep-driving-economic-growth.