Siddharth Singh works in the Office of the Chief Energy Economist of the International Energy Agency (IEA) based in Paris. He is the co‑lead of IEA’s report and workstream on Energy and AI, and a contributing author of the World Energy Outlook. He has previously worked at specialized energy institutions in Berlin, Oslo, and New Delhi. He is the author of The Great Smog of India (Penguin, 2018), a book on India’s air pollution crisis. Singh, a speaker on a June 18 panel on AI’s energy needs moderated by The Innovator’s Editor-in-Chief, agreed to be interviewed on the same topic.
Q: Why has the IEA launched a new report and work stream on Energy and AI?
SS: The IEA published its first major study on the nexus between AI and energy in 2025. We don’t usually do special thematic reports on the same sector or technology in quick succession. However, the AI world is moving very fast, and we wanted to understand the energy implications of the new advancements. With that in mind, we published a new special report on energy and AI a few weeks ago.
Here are a few data points that illustrate how fast things are moving. Capital expenditure of just five technology companies is now larger than global investment in oil and natural gas production. Many jurisdictions are seeing project pipelines accelerate dramatically, although not all projects will come to fruition. Those that are moving forward are doing so at pace: the IEA’s tracking shows that specialized AI data centers that are often called “AI factories” have more than tripled in capacity in the past 18 months.
Secondly, the global electricity demand of data centers grew by 17% in 2025. Electricity consumption from AI-focused data centers grew even faster, surging 50% in 2025. For context, global electricity use grew by 3% during the same period. While there are no comprehensive statistics on the frequency and depth of AI usage around the world, major model providers reported a threefold increase in active users and a fivefold increase in revenue over the past year, highlighting the rapid growth of demand.
Our projection is that electricity consumption from data centers will double by 2030 to close to 1,000 terawatt-hours (TWh), using more electricity than the whole country of Japan – one of the largest economies in the world. As data centers tend to cluster around each other and usually come up in or around urban areas, they can pose challenges to legacy electricity systems.
At the IEA, we wanted to inform our stakeholders of the implications of the rise of AI on energy – including looking at the potential of AI in optimizing the energy sector itself.
Q: What do you see as the major barriers to solving the AI energy conundrum?
Planning and regulatory systems are being stretched by the wave of project applications for data centers, amid a broader trend of rapid load growth and electrification. Social acceptability is also a growing issue, as communities push back against data center projects, and concerns about affordability and environmental impacts rise. Essential elements within the IT industry are currently facing limitations; notably, a shortage of high-bandwidth memory – integral to AI chip production – has developed in recent months.
Q: What percentage of the world’s energy are data centers expected to consume?
SS: Data centers account for about 1.5% of electricity consumption today. This rises to 3% by 2030 in our Base Case. That doesn’t seem like a lot, but the issue is that data centers are not uniformly distributed. Countries that account for 60% of the world’s population have only 9% of data center capacity located within them. Data centers in the U.S., European Union, and eastern China account for a vast majority of data center capacity. In the areas where data centers are clustered, they can consume a much larger percentage of energy. For example, in Ireland and Northern Virginia, both major hubs of data centers, they account for 20-30% of electricity consumption.
Data center operators prefer being connected to the grid. However, because data centers are being built really quickly, the grid can’t always accommodate them, so several data center operators are bypassing the grid with on-site power generation that is mostly fueled by natural gas. Accounting for both on-site and grid-connected emissions from electricity generation for data centers, we estimate that emissions from this sector are set to double by 2030. Of course, tech companies still largely have net-zero decarbonization targets, and so they continue to pour lots of money into renewables, nuclear, and batteries, helping frontier energy technologies to move from the lab to the market. The tech sector is also the single largest procurer of renewable electricity.
Q: The tech companies may have sustainability goals, but they are not transparent about how much energy and water they use, and the way they measure the use is not uniform. Is there a need for a global standard?
SS: There is a need for greater transparency and standard setting. Public concerns over data centers in some regions might encourage more technology companies to release the energy intensity and environmental metrics associated with AI models and data centers alike. This transparency will also help policymakers make better decisions.
Q: What role do you see emerging technologies playing in helping supply green energy to data centers?
SS: In recent years there has been a renewed focus in the development of various energy technologies, such as small modular nuclear reactors (SMR) and advanced geothermal. Many new energy technologies are moving closer to commercialization, but there are still a range of challenges they need to overcome including how quickly they can be scaled up and become cost competitive. Technology companies have been enablers of such new energy technologies. However, we remain cautious and see the deployment of technologies such as nuclear SMR only in the 2030s. And in general, there is a continued need for public support of energy innovation, including in fields such as nuclear fusion.
Q: What percentage of electricity is currently being generated by renewables?
SS: Right now, about 35% of electricity globally is generated by renewable sources such as solar and hydropower. Data centers need electricity round-the-clock. Some technology companies have been working to combine power from renewables, supported by batteries and some computational flexibility, to ensure that data centers get clean power throughout the day on an hourly basis. Such innovations can work to make data centers more sustainable, although it must be noted that such efforts are currently only at a very small scale.
Q: How can new battery technology help introduce more flexibility into the grid?
SS: It is an important part of the solution to getting data centers integrated into the grid. Unlike traditional data center operations, AI training and model inference induce large and rapid power swings, making technologies such as batteries critical in modern data centers. By 2030, around 20-25 GW of battery storage could be installed in data centers globally, potentially making them a grid asset if the incentives are right. Recent months have seen a data center operator sign an agreement for the largest battery project ever by energy capacity, which is four times larger than the previous record-holder, helping to commercialize long-duration energy storage. Sophisticated, grid-interactive on-site power assets, such as battery storage, can help data centers support grid operations, moving data centers from grid loads to grid resources.
Q: What are some other ways that AI has the potential to be an important tool to enhance sustainability?
SS: AI has immense potential to optimize energy operations by unlocking efficiencies, resilience, and innovation. AI technologies monitor grids, transformers, and other energy equipment to reduce unexpected failures and outages, and AI and digital grid-enhancing technologies are key to optimizing the use of existing grid capacity, helping to offset lengthy and costly grid expansions. Improved hyperlocal weather forecasting using AI also helps integrate more renewable power into the grid. These are not just experiments. Many such AI solutions have already been deployed commercially.
Q: In your view, what needs to be done to ensure that the benefits of AI outweigh the negatives when it comes to energy demands?
SS: There are two broad strategies that go hand-in-hand. On the one hand, cleaner electricity for data centers, and on the other, AI applications for energy sector optimization. For both of these to materialize, several barriers will need to be overcome. A reformation of grid-connection queues and powering strategies is vital. And of course, barriers that impede the applications of AI in the energy sector also need to be overcome, including the need for access to energy sector data to train AI models, and regulation that facilitates digitalization and interoperability of energy infrastructure and equipment.
