AI's electricity demand is not the problem. Its inflexibility is
Getty Images
Getty Images· 5 min read
Artificial intelligence is rapidly becoming the latest villain in the energy debate. Data centres consume substantial amounts of electricity, require reliable power around the clock and are often proposed in regions where electricity networks are already congested.
The numbers explain the concern. Global data-centre electricity consumption is expected to more than double to approximately 945 TWh by 2030, slightly more than Japan's total electricity use today. Demand from AI-optimised data centres could increase more than fourfold over the same period (IEA).
But electricity consumption alone is the wrong metric. The real issue is where, when and how that electricity is consumed.
Europe is trying to electrify transport, heating and industry. It would therefore be contradictory to celebrate rising electricity demand in these sectors while treating every additional megawatt used by data centres as a failure.
Electricity demand can support the energy transition. Large and predictable consumers can provide the revenue certainty needed to invest in renewable generation, storage and network infrastructure. Data centres can also strengthen the business case for clean electricity in regions where renewable resources are available but local demand remains limited.
The problem emerges when data centres are built in already congested locations, demand an uninterrupted connection and leave the costs of reinforcing the electricity system to other users.
Ireland demonstrates the scale of this challenge. Data centres accounted for approximately 22% of national electricity demand in 2024 and could reach 31% by 2034. The country has consequently introduced connection requirements under which new data centres must provide onsite or nearby generation equal to their peak demand (ACER).
This is not an argument against data centres. It is an argument for treating their location and operation as part of energy-system planning.
Traditional industrial development usually brings electricity infrastructure to the factory. Data centres offer an opportunity to reverse that logic.
Unlike a steel mill built around a port or a chemical plant integrated into an industrial cluster, many digital activities have greater locational flexibility. New facilities can be directed towards areas with abundant renewable production, available network capacity and useful demand for their residual heat.
Location should therefore become a central sustainability criterion. A data centre built beside constrained renewable generation in northern Sweden or Spain is fundamentally different from one added to an already congested metropolitan electricity system.
Timing matters as well. Not every digital process requires exactly the same level of continuous operation. Some computing tasks can be shifted by minutes or hours, while batteries and backup systems can reduce demand during periods of system stress.
That flexibility has value. ACER is now explicitly calling for system assessments to include demand response from data centres alongside batteries and other non-fossil flexibility resources.
Almost all the electricity entering a data centre eventually becomes heat. In most facilities, that heat is still treated as a disposal problem rather than an energy resource.
It will not be usable everywhere. Low-temperature heat is difficult to transport over long distances, and constructing district-heating infrastructure can be expensive. But where data centres are located close to cities, industry, greenhouses or water-treatment installations, the opportunity becomes significant.
In June 2026, 14 European industry associations agreed to develop a model connecting data-centre operators, energy companies and public authorities, including through the reuse of waste heat in district-heating systems (European Commission).
This should become the norm: assess heat demand, water availability, network capacity and renewable supply before approving the location, not after construction.
Europe does not need to choose between AI development and the energy transition. It needs rules that make the two reinforce one another.
Grid access for major new data centres should increasingly depend on location, flexible operation, onsite storage, renewable procurement, heat recovery and a fair contribution to network expansion. The European Commission's emerging data-centre rating system is useful, but transparency alone will not be enough.
AI will consume more electricity. That is now almost unavoidable. The relevant question is whether Europe passively accommodates that demand or uses it to build a cleaner, more flexible and better-financed electricity system.
The problem is not that data centres use electricity. It is that we still allow them to use it without sufficiently considering when, where and at what cost to the wider system.
illuminem Voices is a democratic space presenting the opinions of leading Sustainability Thought Leaders, their views do not necessarily represent those of illuminem.
The world needs sustainability knowledge. At illuminem, no interest group or shareholder can influence our work. Thank you for supporting our mission to make high-quality and independent sustainability information free for all. Every contribution helps. Thank you for donating today.
illuminem briefings

Nuclear · Public Governance
illuminem briefings

Power Grid · Pollution
illuminem briefings

Oil & Gas · Ethical Governance
Oil Price

Energy Transition · Solar
Associated Press

Oil & Gas · Public Governance
Euronews

Solar · Renewables