AI's Energy Challenges: Efficiency, Sustainability

Artificial Intelligence (AI) is increasingly assisting energy companies with activities such as oil and gas discovery and energy demand optimization. However, it has created sustainability challenges due to its high energy consumption. A recent report by The Economist Intelligence Unit (EIU) highlights Shell’s 2023 collaboration with SparkCognition to enhance subterranean exploration and Cosmo Oil’s use of digital twins to optimize refinery operations as instances where AI technology is being leveraged to improve efficiency and accuracy.

Blanca Martínez, Director of International Cooperation, ASEA, mentions that significant benefits can be achieved through the incorporation of new technologies such as: improvement of personnel safety and operational risk reduction through the enablement of remote operations and control from shore, minimizing the need for on-site interventions. Moreover, standardized and repeatable designs contribute to efficient project delivery, optimizing time and resources.

Hermes Aguirre, Country Vice President, Halliburton, highlights the challenge presented by reservoirs and wells with globally unique characteristics, especially the deepest wells. He notes that although technologies have demonstrated certain limitations, the great opportunities lie in how to assimilate new technological developments and the potential of artificial intelligence to better understand fields. This approach would allow drilling and optimization plans to be more efficiently incorporated to achieve the desired results. He also highlights the scope of digitalization and technology to overcome current barriers.

The strategic integration of sensors, automation, digitalization, and robotics within facilities results in significant advantages. This may not only minimize maintenance needs, leading to improved project economics, but also contribute to a substantial reduction in the overall environmental impact, according to EIU.  This approach can also positively impact regulatory supervision, as the proactive nature of these technologies makes prevention easier to achieve. With automated prevention alerts and immediate emergency notifications, regulators can respond more effectively, ensuring a safer and more sustainable operation. By embracing these technologies, the industry can foster a culture of proactive risk management, further enhancing safety and environmental supervision.

However, the report also emphasizes that sustainability will be a significant challenge for generative AI, given its substantial electricity requirements and carbon footprint. According to the International Energy Agency (IEA), data center electricity demand could double between 2022 and 2026. Combined expansion of traditional and AI data centers, along with chip foundries, will increase demand by 177TWh between 2023 and 2030, reaching a total of 307TWh, more than double the 2023 levels of 130TWh, as reported by MBN. 

Mexico currently has 109 data centers in operation, 21 under construction and another seven announced for the next five years. In terms of energy consumption, data centers in operation consume 111.5MW, those under construction will require 475.7MW and future infrastructure will demand 1,017MW, according to the Mexican Association of Data Centers (MEXDC).  In total, the data center industry in Mexico is projected to require an additional 1,492.7MW over the next five years, positioning the country as the second-largest consumer of electricity for this sector in Latin America. 

Companies like Equinix emphasize the need to invest in renewable energy and establish key initiatives such as optimizing air flow, adjusting operating temperatures, and using liquid cooling to promote the sustainable development of these centers. The United States and the European Union are also implementing regulations to reduce AI’s environmental impact. The EU has introduced the Energy Efficiency Directive, mandating data center operators to report on energy use and carbon emissions starting September 2024.