A solution to AI's energy dependency? Increased AI, according to the IEA.

      The International Energy Agency (IEA) has released its first significant report assessing the AI gold rush's effects on global energy consumption, revealing concerning and potentially conflicting insights.

      The report predicts that energy consumption from data centres, including those used for artificial intelligence, will double in the next five years, reaching 3% of global energy usage. AI-specific energy demands could account for more than half of this increase, according to the findings.

      Currently, some data centres use as much electricity as 100,000 homes, and future hyperscale data centres may consume 20 times this amount, as stated by the IEA. By 2030, it's expected that data centres will derive 50% of their energy from renewable sources, with the rest coming from a mix of coal, nuclear energy, and new natural gas plants.

      While the findings are negative for the climate, the IEA notes a silver lining. Although AI is projected to consume more energy, its potential to enhance efficiencies in power systems and discover new materials might provide a balancing effect.

      Fatih Birol, the IEA’s executive director, remarked, “With the rise of AI, the energy sector is at the forefront of one of the most significant technological revolutions of our time. While AI is a potentially powerful tool, it is up to our societies, governments, and companies to determine how we utilize it.”

      AI can improve power grid efficiency, boost energy production from solar and wind through superior weather forecasting, and identify leaks in critical infrastructure. Additionally, it can aid in planning transportation routes and urban design and may help in finding new sustainable materials for technologies like batteries.

      However, the IEA cautioned that the overall effect of these AI-driven solutions would be “marginal” unless governments create conducive “enabling conditions.” The report stated, “The net effect of AI on emissions—and therefore climate change—will hinge on how AI applications are implemented, what incentives and business models emerge, and how regulatory frameworks adapt to the changing AI landscape.”

      Divisions abound in the discourse surrounding AI and energy. While AI could theoretically reduce energy consumption, significant uncertainties persist, especially since the technology’s adverse climate consequences are already established.

      The IEA estimates that data centres will be responsible for 1.4% of global “combustion emissions” by 2030, nearly tripling the current figure and approaching that of air travel. Although this might seem small, the IEA's estimate does not include the emissions generated from the construction of new data centres and the manufacturing of their materials.

      Alex de Vries, a researcher from VU Amsterdam and founder of Digiconomist, suggested to Nature that the IEA may have underestimated the increase in energy consumption due to AI. “Regardless of the precise number, we’re looking at several percentages of our global electricity consumption,” he stated. He cautioned that the rise in data centre energy use “could pose a significant risk to achieving our climate objectives.”

      Claude Turmes, Luxembourg’s energy minister, criticized the IEA for presenting an overly optimistic perspective, stating that it fails to acknowledge the harsh realities that policymakers must confront. He claimed that instead of offering practical recommendations for regulating and mitigating the significant negative effects of AI and mega data centres on the energy system, the IEA and its director, Fatih Birol, are essentially catering to the new Trump administration and the technology companies that support it.

      Besides AI, there are more established methods to reduce energy consumption in data centres, such as immersion cooling technologies developed by companies like Asperitas from the Netherlands, Submer from Spain, and Iceotope from the UK. Another approach involves repurposing data centre heat for different uses, which is the focus of the UK venture DeepGreen.

      All these innovative solutions will need to rapidly scale if they are to make a significant impact on the energy demands of data centres. Ultimately, it is also crucial to start utilizing computing power more effectively.

      The discussion on sustainable AI will continue at the TNW Conference, scheduled for June 19-20 in Amsterdam. Tickets for the event are currently available, and using the code TNWXMEDIA2025 at checkout will provide a 30% discount.