AI needs power. Power needs AI.
A two-way revolution is underway between AI and electricity. This page marks the start. Our landmark study on Power for AI and AI for Power will be unveiled in 2027
Why now
AI x Power
European electricity demand has not seen any dramatic changes for the past fifteen years. Now it’s moving – driven largely by data centres and the artificial intelligence (AI) workloads they host: large language models, inference at scale, and the hyperscaler and sovereign cloud build-out racing across Europe. For utilities, system operators and the European Union attempting to plan out grid infrastructure and power market design up to 2050 and beyond, this fundamentally changes the math.
Two things are now happening in parallel. AI data centres have become a significant new customer of the electricity sector, while the utilities powering them are deploying the same AI models to do real operational work. The decisions taken today will shape Europe’s electricity grid for the long term.
power for ai
The data-centre boom is the most concentrated, contracted, fastest-ramping new load utilities have seen in a generation. How it gets connected, sited and priced – and how energy efficiency and clean energy sourcing are built into that connection from the start – will shape Europe’s AI sovereignty and the next round of grid investments.
ai for power
Utilities are putting machine learning, deep learning and increasingly generative AI to work across the value chain, from real-time forecasting and dynamic line rating to predictive maintenance and customer service. As the electricity system is getting harder to run while it moves away from fossil fuels and towards variable renewables, artificial intelligence is becoming an essential lever for keeping it stable.
How this plays out will set Europe’s position in the AI economy, how quickly system operators can connect new customers to the electricity grid, and whether data centres end up strengthening long-term clean energy systems or spreading their capacity even thinner.
Power4AI
Power for AI
Demand growth
~28%
of the increase in European electricity demand to 2030 is expected to come from data centres
Forecast range
149 - 287 TWh
forecast range for 2030 European data-centre electricity consumption
Grid connection
7-10 years
current grid-connection backlog in European data-centre hotspots, against a typical 18–24 months to build a data centre
1
AI sovereignty starts with the power system
Whether Europe ends up hosting frontier AI, or just renting it from elsewhere, will be decided largely by the power it can offer: how much, at what price, with what reliability. “This underscores the strategic importance of power infrastructure in delivering Europe’s cutting-edge technologies — placing electricity policy, at the heart of the industrial and security debate” the Draghi report opened. The conversation now sits there, and it is an opportunity for the sector to shape it.
149 - 287 TWh
forecast range for 2030 European data-centre electricity consumption
Source: Eurelectric Power Barometer 2025
2
This load is unlike anything utilities have planned for
Today’s AI data centres bear little resemblance to the colocation sites of the 2010s. They’re bigger, far denser, ramp up faster, and increasingly turn up in places nobody had on the planning radar. If you’re running a transmission plan or a regional distribution system operator (DSO) investment cycle, the old model breaks.
Scale. Electricity demand in Europe has been broadly flat since 2010. Yet by 2030, European data-centre demand will land somewhere between 149 and 287 TWh. For DSOs and transmission system operators (TSOs), this is the first multi-year, capital-backed demand pipeline the sector has seen since the post-war industrial era.
Geography. FLAP-D (Frankfurt, London, Amsterdam, Paris, Dublin) still holds most of Europe’s capacity, but concentration is extreme: 87% of Ireland’s data centres sit in a single city. Such concentration creates resilience and security exposure that a broader rollout can help address. Growth is now spilling into Tier 2 and 3 cities and the Nordics, where cheap, clean power and available land attract investment on their own.
Climate is also tilting siting decisions. Investors increasingly favour geographies whose ambient conditions — cooler temperatures, stable humidity, abundant cold water — do part of the cooling work. That makes siting a cohesion-policy question for Brussels as much as a market one.
Shape. Global data-centre workloads accelerate from 8% annual growth to 14% in 2025-2030. AI inference alone goes from 9 GW to 74 GW — a 52% compound annual growth rate. EMEA capacity expands from 21 GW to 34 GW by 2030.
European final power demand growth (TwH), 2023 to 2030 expected
Figure 1 – Sources: Eurelectric Data Centre Stocktake 2026, EY-Parthenon analysis of Goldman Sachs data, EY electricity and resources transition acceleration model. Europe = EU27 + United Kingdom.
3
Speed-to-power, not power itself, is the binding constraint
Europe has no shortage of electricity in the near term. But it is short of time. Connection queues are getting longer; “ghost” requests, the speculative projects looking for buyers and that never get built, inflate statistics on expected production increase; and high-voltage projects routinely take five to ten years to move from application to delivery. The danger is not that AI investments go elsewhere because Europe has no power. It is that AI investment goes elsewhere because Europe can’t connect it in time.
7 - 10 years
Source: Elisabeth Cremona & Pawel Czyzak (2025). Grids for data centres: ambitious grid planning can win Europe’s AI race.
4
Clean, stable power is Europe's long game
Past the connection bottleneck, the longer-run question is the kind of system Europe ends up building. More carbon free generation, more storage, more flexibility, and a grid that can actually move what is added. Whether and how Europe succeeds in doing that affordably is the key challenge for the rest of the decade.
5
How the roll-out happens decides whether data centres help or hurt
Done right, AI demand is a net positive for the electricity system. The biggest effect is on infrastructure costs in the electricity bills: every kWh of demand we spread across a bigger customer base is a bit less burden on households and SMEs. Long-dated contracts from hyperscalers also help utilities build a credible investment case with the regulator.
Get it wrong, and the dynamic reverses. Data centres get framed as cost drivers; mayors and ministers turn against them; permitting tightens; capacity is capped or paused. The same demand wedge that could underwrite the next decade of grid investment ends up being the political problem of the decade instead.
European data centre overview
FLAP-D hubs Constrained and policy-managed
Frankfurt, London, Amsterdam, Paris, Dublin
- Largest data centre markets in Europe
- Sustained high demand (100 to 250 MW/yr)
Nordics
Norway, Sweden, Finland, Denmark, Iceland
- Strategic growth markets thanks to low costs of electricity, site availability and conditions
- Moderate to high demand (20 to 100 MW/yr)
Tier 2 European markets
Attractive but infrastructure limited
Madrid, Berlin, Milan, Rome, Lisbon
- Increasing interest from the major cloud players in recent years
- Increasing demand as cloud players enter (30 to 100 MW/yr)
Tier 3 European markets
Emerging with lower immediacy
Barcelona, Vienna, Marseille, Brussels, Warsaw
- Secondary markets in countries with larger data centre hubs or primary markets in countries with lower demand
- Low, less predictable demand (0 to 30 MW/yr)
Figure 2: Sources: EY-Parthenon Data Centre Benchmarks Database, EY-Parthenon analysis
Twin Transition Commitments
The Twin Transition Commitments: utilities and hyperscalers acting together
A build-out at this scale cannot be coordinated alone. That is why, at Power Summit 2026, the European power sector and the digital sector signed the Twin Transition Commitments, developed in partnership with EY-Parthenon.
Over the next months, the two sides have agreed to work together on practical pathways for the timely, efficient and sustainable scaling of data centres in Europe, and on how to plug them properly into the electricity system. This work will feed into Eurelectric’s Era of Electric Intelligence flagship study being prepared for 2027.
Over the next 6 months we will jointly explore:
Three research areas
1. Data centres development:
demand and priority locations, flexibility and pathways to mitigate overall system costs.
2. Power system readiness:
how Europe can meet AI-driven demand through grid expansion, advanced technologies, and improving demand forecasting.
3. Sustainable & secure strategies:
utility–hyperscaler collaboration models, new contractual arrangements, energy efficiency and clean electricity sourcing.
Three policy and market enablers
1. Connections & stability:
flexible-connection frameworks and queue reforms; data centres contribution to grid stability; maximizing the use of existing and new grids.
2. Market incentives & affordability:
advancing carbon-free power procurement, PPAs and system-efficient cost allocation for data centre integration.
3. Coordination & planning strategies:
streamlining permitting for new power capacity; long-term siting and planning coordination with data centres.
Twin Commitments signatories: Eurelectric, Fortum, EDP, Aelec, A2A, PPC, Engie, Enedis, ESB, Google, Schneider Electric, Sepia infrastructure, GE Vernova, Iron Mountain data centers, CTC Global, Kaluza, CGI, EPRI, Siemens Smart Infrastructure, Enline, Hyosung Heavy Industries, Landys+Gyr.
Power for AI case studies
Case Studies
Six examples showing how grid capacity, advanced conductors and AI-enabled monitoring can support faster data centre connection and more flexible use of existing infrastructure.
CTC Global: Grid with integrated AI-capable capacity and integrity monitoring
CTC Global: Rapid doubling of energy supply to data centre in Mons, Belgium thanks to ACCC® Conductor technology
Iberdrola and Echelon create a joint venture to develop data centres in Spain
Siemens and Delta power solutions cut data center deployment time, costs, and carbon emissions
Schneider Electric and Start Campus Establish Scalable, Sustainable Foundation for AI and Cloud Infrastructure in Portugal
Microsoft Data centers provide 40% of Fortum’s electricity-based district heating in Espoo, Finland.
AI4Power
AI for Power
Run the camera the other way, and AI is doing real work inside utilities, too. Machine learning, deep learning and, increasingly, generative AI are moving from R&D pilots into day-to-day operations, in a system that gets harder to run every year as it adds variable generation like wind, solar, and other distributed assets like EVs and heat pumps.
These deployments cut across the entire value chain – from generation and grid operations through trading, market participation and customer service – with no single ‘home’ for AI inside utilities. The catalogue below pulls together specific cases from across Europe, what they do, how they were built, and what they deliver, so utilities can learn from each other.
AI Catalogue
AI for Power use case catalogue
Explore how leading utilities are unlocking the potential of AI to solve real-world challenges, drive operational excellence, and accelerate the transition to a more digital, flexible, and sustainable energy future. Our AI for Power use case catalogue was developed in partnership with EY-Parthenon, drawing on case studies from leading utilities and technology providers across Europe.
Why explore AI for Power?
The power sector is evolving – becoming smarter, more digital, and more interconnected. Artificial intelligence is at the centre of this change, helping utilities and innovators optimise assets, enhance forecasting, and make faster, more informed decisions. To realise the full benefits of AI, we need a practical, shared view of what’s working and where value is being delivered.
What’s inside the catalogue?
Explore a curated collection of real-world AI case studies from utilities and technology leaders, showcasing how AI is already making a difference across the power value chain – from generation and networks to markets and customer solutions. Each case highlights practical applications, business results, and lessons learned, showing how AI is moving from pilot projects to core operations.
How to use this catalogue?
Navigate the catalogue using filters to quickly find case studies by value chain, technology, or business goal. Compare maturity and impact, explore detailed case studies, and discover approaches you can adapt for your own organisation. Whether you’re benchmarking, learning, or connecting, this catalogue is your starting point for AI in power.
What the real-world AI case studies reveal about how the power sector is adopting AI
AI adoption in the power sector is evolving from prediction to autonomous system orchestration
~70%
case studies leverage
machine learning
~23%
deep learning
~23%
case studies leverage
generative AI
~9%
case studies leverage
agentic AI
Across the ~50 case studies, AI adoption in the power sector follows a clear progression, with each step below maturing steadily. The majority of applications are still focused on prediction and optimisation, but utilities are increasingly adopting more advanced capabilities to interpret complex systems, support decision-making, and begin coordinating operations in real time. A shift from isolated use cases towards a more integrated, system-level intelligence.
Real-world AI case studies reveal 5 trends about how the power sector is adopting AI.
1. AI is deployed where operation complexity is highest
AI adoption is driven by the complexity of decisions rather than just data availability.
2. Grid edge is becoming the primary AI frontier with customer as focus
The majority of AI case studies are deployed at the grid edge, with a shift towards decentralisation – and AI is following this shift.
3. Most AI use cases are locally deployed and remain highly context-specific
AI deployment remains localised, with scaled case studies focused on cost optimisation.
4. AI deployment is increasingly ecosystem-driven
AI solutions are largely co-developed or integrated through external partners and vendors.
5. AI adoption is increasingly driven by data and integration capabilities
Successful AI deployment depends on strong data foundations and seamless integration.
AI for power case studies
Electric Intelligence Catalogue
These case studies are a starting point. The full picture – on data centre integration, grid readiness, and AI deployment across Europe’s electricity system – will emerge in Eurelectric’s Era of Electric Intelligence flagship study in 2027.
Do you want to see your AI application in the energy sector be published on this website? Contact our business development team at business@eurelectric.org.
Do you want to join the Twin Transition Commitments and help us shape our new study? Reach out to nsteinwand@eurelectric.org.