May 2025 issue of the Oxford Energy Forum looks at how AI is changing the electricity systems globally.

What: Experts explore how AI is increasing power use, influencing how electricity grids are run, and possibly reshaping how energy markets work.

Why: The energy sector now sits at the intersection of two powerful forces: the clean energy transition and the AI revolution.

What next: AI might help make electricity systems more flexible, reliable and green.

The rapid growth of Artificial Intelligence (AI) is transforming how industries operate, particularly the electricity sector. As AI becomes more powerful, the demand for electricity, especially from data centres that house AI systems, is rising fast. This demand comes at a critical time, as the world is trying to shift to cleaner energy. The energy sector now sits at the intersection of two powerful forces: the clean energy transition and the AI revolution.

The latest edition of the Oxford Energy Forum (OEF), a quarterly journal for debating energy issues and policies by the Oxford Institute for Energy Studies, looks at how AI is changing the electricity system globally. Experts explore how AI is increasing power use, influencing how grids are run, and possibly reshaping how energy markets work. They also highlight how AI might help make electricity systems more flexible, reliable and green. The May 2025 issue of the OEF brings together global and regional insights, policy discussions and technological solutions, providing a comprehensive overview of this major development in modern energy systems.

Global trends and risks

In the opening chapter, Plutarco Naranjo and David Robinson explain that geopolitical competition is accelerating AI development, potentially increasing electricity demand sixfold by 2030. While improvements in hardware and data centre management can enhance system efficiency, the inference stage of AI, where trained models generate results, still consumes a significant amount of electricity. They warn that this global AI race might delay the transition to clean energy by promoting greater fossil fuel use. To address this issue, they recommend implementing policies that link AI electricity demand to market signals rather than subsidies, and they advocate international cooperation for sustainable AI development.

Next, Charlie Wilson, Yee Van Fan and Felippa Amanta argue that AI’s indirect effects on energy use might be even greater than its direct ones. While data centres will increase electricity use, the real concern is how AI changes energy use across all sectors. Some uses save energy; others do the opposite. These indirect effects are hard to regulate, so the authors suggest using proportionality tests for AI projects and tracking their broader impacts.

Operational transformation and market impacts

Sam Young highlights three key ways AI will change how energy systems operate: 1) enabling coordination, where AI can link many small energy sources into efficient systems; 2) enhancing market optimisation to improve trading strategies and also expose flaws in market rules; and 3) introducing systemic risks where competing AI algorithms might behave in ways that cause disruptions during high-stress events. Young urges the industry to rethink how it manages the energy system, focusing more on edge devices, dynamic operations and clear goals for AI use.

With a focus on the US, Michael Hochberg shows how AI is driving a sharp increase in electricity use, as data centres could consume 12% of total US electricity by 2030, especially in Texas and the East Coast. While this opens opportunities for innovation, it also raises issues like grid reliability, retiring old power plants, and rising emissions. Hochberg questions whether AI’s efficiency gains can keep up with its fast-growing power needs.

In the UK, Malcolm Keay points to similar concerns. The UK wants to lead in AI and reform its electricity market while cutting emissions, but these goals clash. According to Keay, it is hard to predict AI’s power demand, high UK electricity prices affect AI competitiveness, and AI infrastructure doesn’t always fit with planned reforms like zonal pricing. This makes it tough for the UK to balance leadership in AI with clean energy goals.

European governance and regulation

From a European perspective, Irene Niet and Rinie van Est stress the need to watch AI’s wider social and environmental costs. Although AI can improve energy efficiency, its infrastructure is energy-intensive, leading to more carbon emissions. They also warn that AI might give Big Tech companies too much control over energy systems, limiting democratic oversight. They call for strong governance frameworks through laws like the European AI Act.

Expanding on this, Eva de Winkel, Betül Mamudi, Roel Dobbe, and Jochen Cremer focus on electrical distribution grids. AI can help manage congestion and improve grid forecasting, but its use in vital infrastructure poses safety and ethical risks. The European AI Act treats these systems as high-risk, requiring clear rules on transparency, risk management and human oversight. The authors propose five key governance questions and suggest using safety frameworks to guide AI adoption.

Technical innovations and energy resource management

Alejandro D. Domínguez-García, Hanchen Xu and Dimitra Apostolopoulou explain that traditional grid control methods are outdated. As distributed energy resources (DERs) like rooftop solar increase, managing power flows becomes more complex. They propose using machine learning and real-time data to improve reliability and flexibility. They further showcase AI-based solutions for voltage regulation and DER coordination but note that cybersecurity, accuracy, and environmental impacts remain concerns.

Alex Papalexopoulos and Mayank Saxena focus on virtual power plants, where AI helps bundle DERs for market participation. Their platform uses AI and machine learning (ML) to improve grid balance and energy market strategies. If regulations and technology improve, this approach could reduce reliance on traditional fossil-fuel power stations.

Utility operations and cybersecurity

From a utility perspective, Rafael San Juan Moya sees AI as vital in handling growing demand and complexity. AI can optimise planning and investments, enable automated fault detection, improve renewable energy forecasting, cut costs through predictive maintenance and enhance customer service. However, San Juan Moya warns, careful planning is needed to manage the energy demands of AI systems, and cybersecurity must be prioritised.

On that topic, Xiang Huo, Justin Leiden, Emily Payne, Astrid Layton and Katherine R. Davis examine AI’s double-edged role in cybersecurity. AI can protect grid systems from cyberattacks, but it can also introduce new risks. They advocate for rigorous testing, clear policies, and collaboration among governments, industry, and academia to ensure AI is used responsibly.

Data centres: from challenge to solution

Line Roald argues that while data centres are heavy electricity users, they could also support grid operations. By making their computing loads more flexible, data centres must adjust their usage to the needs of the electricity grid, just like renewable energy providers have done. Roald suggests ways of controlling power spikes and adjusting computing workloads to make data centres more responsive.

Dimitra Apostolopoulou and Rahmat Poudineh explore how waste heat recovery from data centres could help improve energy efficiency. But this depends on policy support, infrastructure and investment. Countries like Denmark and Sweden have done well with this, while others lag owing to a lack of incentives and weak district heating networks.

Small modular reactors and the AI power race

Alvaro J. Lopez-Lopez considers whether small modular reactors (SMRs) could supply the rising energy needs of AI data centres. These reactors are low-carbon, reliable and compact, making them good candidates. However, they are costly, produce nuclear waste, face safety and approval challenges, and are not yet widely deployed. Still, tech giants like Amazon, Google and Microsoft are investing in SMRs, and countries like China and Russia are leading in deployment.

Conclusion

Together, these contributions paint a detailed picture of how AI is becoming deeply linked to electricity systems. As AI continues to expand, managing its energy impact will require clever policies, technical innovation and strong collaboration across sectors and countries. Balancing AI growth with clean, secure, and reliable energy is now a global priority, the authors conclude.