The global race to develop artificial intelligence is also becoming a race to secure electricity. New data centres need large volumes of reliable power, often in locations where the network is already dealing with renewable connections, industrial electrification and ageing infrastructure.
For technology companies, electricity is no longer only an operating expense. It can determine where a project is built, how quickly it becomes operational and whether the original investment schedule remains commercially viable.
The International Energy Agency expects global data centre electricity consumption to rise from approximately 485 TWh in 2025 to around 950 TWh in 2030. That would place data centres close to 3% of total global electricity demand. AI-focused facilities are expected to expand even faster, with their consumption potentially tripling over the same five-year period.
The global total tells only part of the story
A near doubling of electricity consumption attracts attention, but the geographic concentration of data centres may be even more important.
Digital infrastructure tends to cluster around a limited number of cities and regions with strong fibre connectivity, access to customers, available land and an established technology ecosystem. Electricity demand therefore does not increase evenly across an entire country. It may rise sharply around a small group of substations that are already approaching their technical limits.
The situation is visible in markets such as Ireland, where data centres account for more than 20% of electricity consumption. Other European countries may have a lower national share, yet still experience serious local constraints around major data centre hubs.
This changes the meaning of power availability. A country may produce enough electricity over the course of a year and still be unable to connect a new facility at the required location. The network must deliver the requested capacity through the correct substation, at the correct voltage and with the level of reliability expected by a mission-critical facility.
Data centres and power infrastructure operate on different timelines
A new data centre can be designed and constructed within several years. Major transmission lines and substations usually take longer because they require network studies, permits, land rights, equipment procurement and extensive construction work.
This difference has made time to power one of the most important measures of data centre viability. Suitable land and an apparently favourable electricity price have limited value if the connection will not be available until long after the planned opening date.
The pressure to secure power quickly can create an uncomfortable consequence. Where networks and clean generation cannot expand fast enough, conventional gas generation may be proposed as the most immediate solution. Existing fossil-fuel plants may also remain in operation for longer than previously expected.
Renewables are expected to supply a large share of the additional electricity required by data centres, but they will not cover the entire increase on their own. The speed and location of new demand will influence how much of the remaining supply comes from low-carbon firm generation and how much comes from coal or natural gas.
AI is not automatically incompatible with decarbonisation. Its infrastructure must, however, be developed as part of the energy transition rather than placed alongside it as a separate policy priority.
An annual renewable contract does not describe hourly reality
Many technology companies procure renewable electricity through power purchase agreements or certificates. These arrangements have supported significant investment in clean generation, but annual accounting can conceal what happens during individual hours.
A data centre may purchase an amount of renewable electricity equal to its yearly consumption while still drawing heavily from the grid during evening peaks or periods of weak wind and solar production. At those moments, the marginal source of electricity may be a gas- or coal-fired plant.
Storage and flexible operation can improve the alignment between consumption and clean generation. They cannot make electricity low-carbon by themselves. A battery charged during a fossil-intensive period simply moves that electricity to a later hour.
The source, timing and additionality of the energy therefore matter. Additionality means that the data centre contributes to new clean generation rather than relying entirely on capacity already used by households, businesses and other industries.
The distinction is more than a question of environmental reporting. It affects local prices, network congestion and the amount of generation capacity the system must maintain.
A new way to assess data centre projects
Traditional data centre assessments concentrate on energy efficiency inside the facility. Metrics such as power usage effectiveness remain useful, but they reveal little about the project’s interaction with the wider electricity system.
Two facilities can consume the same amount of electricity in a year and create very different demands on the network. One may reach maximum consumption during an evening peak and maintain the same load during a grid emergency. The other may use storage, postpone non-urgent computing tasks and limit its electricity import when the system is constrained.
The second project does not necessarily consume less energy. It consumes energy in a way that is easier for the network to accommodate.
| Conventional project question | More useful system-level question |
| How much electricity will the facility consume annually? | What is its maximum grid import, and when will it occur? |
| Has the operator purchased renewable electricity? | Is the electricity matched to consumption by time and supported by new clean generation? |
| Does the site have emergency power? | Can its batteries also reduce peaks or respond to grid events without compromising resilience? |
| What connection capacity has been requested? | Can the project operate under a phased or flexible connection? |
| How efficient is the building? | How will the complete facility affect the local power system? |
This broader assessment does not remove the need for a reliable connection. Data centres remain continuous industrial loads, and many of their services cannot be interrupted. It does show where a project can offer measurable flexibility instead of behaving as an entirely passive consumer.
Europe is beginning to connect its digital and energy ambitions
The European Union wants to expand its AI computing capacity while maintaining its decarbonisation commitments. Until recently, these goals were often discussed through separate digital and energy policies.
That division is becoming difficult to sustain. Data centres currently account for an estimated 2.5% of EU electricity consumption. Under existing policies, the IEA expects them to represent approximately 10% of the growth in EU electricity demand to 2030.
In June 2026, the European Commission published its Strategic Roadmap for Digitalisation and AI in the Energy Sector. The document gives greater attention to demand flexibility, grid optimisation and the sustainable integration of data centres.
The policy direction is important, but practical implementation will take place at project level. Grid operators need reliable technical data. Developers need realistic connection schedules. Energy companies need long-term demand commitments before they invest in new generation or storage.
Faster grid access may be possible when a project can demonstrate that it will limit demand during constrained periods. Such access should be based on contractual obligations and verified performance, not on preferential treatment for large technology companies.
The strongest projects will combine several energy solutions
There is no single technology capable of supplying every AI data centre with uninterrupted low-carbon electricity. Solar and wind generation are variable. Lithium-ion batteries are well suited to short-duration balancing but cannot cover prolonged periods of weak renewable output. Nuclear, hydroelectric and other firm sources depend on the conditions of the individual market.
The most resilient projects are likely to combine several elements: a dependable grid connection, additional clean generation, short- and longer-duration storage, flexible computing and procurement contracts that reflect the hourly energy profile.
These elements should be considered during site selection, not added after the location and requested capacity have already been fixed. The available connection, network condition, distance from generation and possibility of future expansion can change the economics of the entire development.
Land with an uncertain power schedule is not genuinely powered land, regardless of the capacity stated in an early-stage document.
Why does this matter to Power Loop readers?
Power Loop readers include data centre developers, infrastructure investors, energy companies and businesses evaluating opportunities across Central and Eastern Europe. For them, the rise in AI electricity demand is not an abstract global forecast. It directly affects project pipelines, site values, connection queues and investment risk.
The central lesson is that power can no longer be treated as one item in a data centre checklist. Its origin, location, reliability and delivery schedule shape the feasibility of the project from the beginning.
At Power Loop, we see the strongest development opportunities emerging where digital and energy infrastructure are planned together. A site becomes more valuable when its power strategy is technically credible, expandable and compatible with the future requirements of the network.
The AI boom will create new electricity demand regardless of how individual markets respond. The competitive advantage will belong to regions and projects that turn that demand into long-term investment in cleaner generation, stronger grids and useful flexibility.
Sources of information:
- International Energy Agency, Key Questions on Energy and AI – Global electricity demand projections for data centres.
- International Energy Agency, Energy Demand from AI – Regional demand growth and AI-related electricity consumption.
- Vienna University of Economics and Business – Time-to-power constraints and the relationship between AI and energy storage.
- European Commission, Digitalisation and AI in the Energy System – EU policy, demand flexibility and sustainable data centre integration.
