Chair’s Summary of the High-Level Roundtable on Energy & AI

Context: Artificial intelligence (AI) is emerging as one of the most consequential technologies of our time. Surging demand for digital services, including AI, is creating a wave of investment in data centres, raising concern about the challenge of meeting the energy needs of AI (Energy for AI). At the same time, applications of AI in the energy sector could offer numerous benefits, including enhancing efficiency, modernising and improving grid flexibility, accelerating the deployment of clean energy technologies, promoting decarbonisation, and boosting innovation (AI for Energy). The AI revolution is still at an early stage and how it will unfold remains uncertain. To ensure that AI helps to meet energy sector goals, and to manage the energy sector challenges arising from the rapid uptake of AI, there is an urgent need for deeper and more structured dialogue between all relevant stakeholders, including policy makers, the energy industry, the tech sector, financial institutions, and academia and civil society.

Event: The International Energy Agency (IEA) convened the Global Conference on Energy & AI on 4 and 5 December 2024 in Paris, France. The Conference represented a first-of-its-kind event bringing together high-level representatives from government, the energy industry, the tech sector, financial institutions, and civil society and academia. The Conference was structured around the two interlinked themes of Energy for AI and AI for Energy.

Background to the IEA’s work: In 2017, the International Energy Agency published a Special Report on Digitalisation and Energy. Since then, the IEA has expanded its analytical and modelling capacities, data collection, and policy recommendations in this field. In 2024, the IEA Secretariat developed a dedicated workstream on Energy for AI and AI for Energy, which will include a special report on energy and AI in the first half of 2025.

Chair’s summary: It is in this context that the IEA convened the Global Conference on Energy & AI, which culminated in the High-Level Roundtable on Energy and AI on December 5, 2024. The Roundtable was chaired by IEA Executive Director, Dr. Fatih Birol.*

The need for enhanced dialogue: Participants recognised the growing interlinkages between energy and AI and the numerous but still uncertain implications of AI for the energy sector, which will require deeper analysis and understanding in order to shape effective policies and strategies. They thanked the IEA for hosting this event and welcomed the unique opportunity offered by the Global Conference on Energy & AI to engage in a dialogue between policy makers, the energy industry, the tech sector, financial institutions, and academia and civil society.

Participants welcomed the IEA’s work on Energy for AI and AI for Energy and the forthcoming publication of a dedicated report on this topic. In the lead-up to its 2025 G7 Presidency, Canada has welcomed the IEA's display of thought leadership by convening an event gathering decision-makers from government and industry, as well as subject matter experts and scientists. The outcomes of this event will undoubtedly lay the foundational work for Canada's G7 agenda. Canada communicated that it would work with the IEA on an approach that would be formalised in 2025 under Canada's G7 Presidency, and that would have the IEA as well as G7+ and industry partners work towards a more responsible, resilient, and secure response to energy for AI concerns and harnessing of AI for energy. As the host of the upcoming AI Action Summit, France underlined its intention to work with the IEA Secretariat in shaping a robust outcome on the energy and AI nexus. Korea noted that its upcoming hosting of the APEC Energy Ministerial, Clean Energy Ministerial and Mission Innovation Ministerial will provide further opportunities to deepen international dialogue on the intersection of energy and AI and expressed its intention to work with the IEA to continue to advance key discussions on the topic.

Ensuring AI capabilities lead to positive outcomes for the energy sector: Participants highlighted the many challenges faced by the energy sector, from persistent concerns about energy security and affordability to lagging progress towards energy access goals, still-rising greenhouse gas emissions and the insufficient pace of clean energy innovation. Participants recognised that AI could accelerate solutions in the energy sector, provided its potential is better understood and mapped, necessary enabling conditions are identified, its environmental footprint does not negate its positive impacts, countries prioritise research and development to spur innovation in this area, and policies and deployment models are designed with a view towards addressing these challenges. Participants also discussed the need for a skilled workforce in both energy and AI.

Inclusive AI in the energy sector: Participants returned repeatedly to the importance of an inclusive deployment of AI in the energy sector that avoids exacerbating today’s existing digital divides; safeguards the interests of consumers; recognises the challenges of some emerging and developing economies in deploying the technology due to a lack of robust, reliable energy infrastructure; acknowledges the obstacles these countries may face in accessing high-quality data to train and deploy models; and supports the goals of just, orderly and equitable energy transitions. They welcomed the IEA Secretariat’s focus on how AI can be developed and applied to address the specific challenges in emerging and developing economies.

Energy for AI: Participants noted that data centres and the Information and Communication Technology (ICT) sector today make up a small share of global electricity demand. With the rapid growth in digital services and AI, participants highlighted the expectation that electricity demand for data centres and digital services would likely rise substantially in the next few years, despite continued improvements in the efficiency of hardware and software. Participants noted that while data centre electricity demand is expected to remain relatively small in the context of the global energy system, the rapid growth of data centres and their spatial concentration is already putting stress on the planning, permitting and deployment of grid and generation assets at the local level in several jurisdictions. They noted that there are important economic and environmental issues at stake, given that the availability of energy is a key determinant in where data centres are built. Participants noted that the speed of data centre development is creating challenges for infrastructure and regulatory systems, and that it is misaligned with the deployment timelines for energy infrastructure, which are typically longer. They stressed the need for streamlined planning and speedy permitting for energy infrastructure to service demand from data centres. Participants highlighted the importance of addressing the climate and other environmental impacts of data centres, notably water and e-waste, including through a lifecycle perspective. Participants also emphasized the role of nuclear and geothermal energy alongside other renewables in meeting electricity demand from data centres, and noted the role of waste heat capturing for more efficient use of energy.

Improving the understanding of the energy footprint of data centres today and in the future: Participants highlighted the high degree of uncertainty related to the current level and outlook for electricity demand from data centres and highlighted several data gaps that inhibit projections. They stressed the importance of efforts to deepen dialogue, enhance alignment on definitions, promote data collection, and facilitate the sharing of data, regulatory and technical best practices, which would improve understanding of the load, operational characteristics and energy consumption of data centres, both today and in forward-looking scenarios. They welcomed the suggestion that the IEA work with interested parties on a voluntary basis in this regard, including through the establishment of an Observatory on AI, Data Centres and Secure and Innovative Energy Transitions. The Chair of France’s AI Action Summit offered the Summit in February as an opportunity to take stock of these efforts.

Meeting data centre electricity demand sustainably: Participants stressed the importance of improving planning, permitting and infrastructure deployment to sustainably meet the energy needs of data centres going forward, including the need for enhanced coordination across the energy and tech value chains and at various levels of government. They noted the efforts of major tech companies to procure clean energy and invest in early-stage and innovative technologies and supported further efforts in this regard. They also stressed the importance of frameworks and incentives to support continuous efficiency improvements in data centres; ease pressures on power grids by identifying and exploiting the potential to locate and operate data centres at locations where congestion is lower and clean electricity supply is higher; and flexibly allocate data centre workloads across time and space. They called for further work to better identify what opportunities exist and understand the necessary enabling conditions. They also highlighted the potential for AI to support grid modernisation and stronger demand response. They highlighted the role that electricity demand from data centres can play in driving the deployment of new technologies, including conventional and advanced nuclear reactor designs, advanced geothermal, and long-duration energy storage. Additionally, they noted the important role being played by renewables in providing low-emissions electricity for grids and individual consumers, as well as the need for continued policy support to accelerate project development and execution.

AI and energy innovation: Participants highlighted the significant potential for AI to accelerate energy innovation, notably for clean energy technologies, and highlighted several promising examples in this regard, including AI-enabled innovations in battery technology and materials science. At the same time, they noted the need to better understand this potential, including by determining the most promising use cases by technology and sector and the necessary conditions to unlock it, including government-industry cooperation. Participants highlighted that early-stage innovation in novel energy technologies could be substantially accelerated by AI, but that significant barriers to commercialisation and market uptake still need to be overcome. Participants noted with interest the IEA’s proposal to develop a roadmap for accelerating AI-enabled energy technology innovation.

AI and energy optimisation: Participants discussed the numerous beneficial ways that AI is already being deployed in the energy sector today, from forecasting electricity demand and renewables to managing power flows on electricity networks, activating and managing automated demand response, and improving the safety and productivity of energy supply. They highlighted the substantial potential to go further in this regard but noted the barriers to broader adoption of AI tools, including the need for high quality, extensive and structured data, a gap in digital skills, concerns over reliability and security, and the need for fit-for-purpose regulatory frameworks, institutions and processes. Participants called for a more extensive mapping of AI applications for optimising energy systems, which would facilitate faster sharing of lessons and best practices, plus a better understanding of impacts and benefits.

Principles for energy and AI: In his closing remarks, IEA Executive Director Birol highlighted six principles emerging from the discussion, including:

• There is no AI without energy – specifically electricity. The need to ensure that the energy sector rapidly delivers sustainable energy for AI, data centres and digital infrastructure, and that data centres and AI contribute to secure transitions in the electricity sector.
• The critical importance of ensuring that AI deployment in the energy sector contributes to enhancing energy security, including through cybersecurity and by potentially improving grid flexibility.
• The need for greater transparency through trusted sharing of data to facilitate informed decision-making and planning. This includes but is not limited to key metrics on data centre capacity, electricity consumption, procurement and efficiency. 
• The need for inclusive and consumer-centric approaches to deploying AI in the energy sector, in ways that
• don’t exacerbate existing digital and energy access divides and support just, equitable and orderly transitions.
• The importance of delivering enabling and supportive policies and business models to ensure that the adoption of AI delivers secure, affordable and inclusive energy transitions, including by accelerating clean energy innovation and deployment and optimising the performance of today’s energy system.
• The need for continued dialogue between policy makers, the energy industry, the tech sector, financial institutions, and academia and civil society, with international fora such as the IEA playing an important role.

Next steps: Participants looked forward to future opportunities for engagement on this topic and welcomed upcoming processes offering platforms to continue discussions, including the AI Action Summit, the G7/G20, and future IEA convenings.

* This Chair’s Summary of the High-Level Roundtable on Energy & AI is provided under the responsibility of the IEA and represents its own high-level summary of the discussions that took place. It does necessarily reflect the views of all participants.