Skip to content

Project Development

AI Is Creating a New Class of Energy and Infrastructure Demand

The data centre is no longer only a digital or real estate asset. At large scale, it is a power, grid, cooling, water, land and resilience project that must be developed as an integrated infrastructure system.

MZA Consultancy · 15 May 2025 · 9 min read

Artificial intelligence is often discussed as software, computing capability and investment in advanced chips. Its expansion is increasingly constrained by physical infrastructure.

The International Energy Agency estimates that data centres consumed around 415 terawatt hours of electricity in 2024, about 1.5 percent of global electricity use. In its April 2025 outlook, the Agency projects that consumption could more than double to around 945 terawatt hours by 2030. AI is the largest driver of that increase.

The global share remains modest compared with industry, buildings and transport. The development challenge is concentration. Data centres cluster in specific locations, and individual projects can request very large connections within a short timetable. A grid that is adequate at national level can still be constrained around one city or industrial corridor.

The grid connection is part of the project concept

For conventional commercial developments, electricity connection is often treated as a utility workstream that follows site selection. For a large AI data centre, it may determine whether the site is viable at all.

The technology sector can plan and construct a facility faster than the electricity system can approve and build major substations, transmission lines or generation assets. The IEA notes that data centres can be operational in two to three years, while energy infrastructure generally requires longer planning and construction periods. It also estimates that a material share of planned data centre projects could face delay if grid constraints are not addressed.

Developers should therefore secure a credible power pathway before committing to a full programme. That requires clarity on connection capacity, reinforcement works, cost allocation, energisation dates, redundancy and the conditions attached to large new loads. A grid that is adequate at national level can still be constrained around one city, one substation or one industrial corridor.

Power supply is becoming a portfolio decision

A data centre requires continuous, high-quality electricity. Annual renewable procurement can support corporate climate objectives, but it does not by itself guarantee that power will be available at every hour or at the required location.

The IEA expects renewables to meet a large share of the growth in data centre electricity demand, supported by grids and storage. Natural gas, nuclear and other dispatchable sources are also expected to contribute. The practical mix will vary by market.

Project sponsors need to decide how grid supply, long-term power contracts, dedicated generation, batteries, backup systems and demand management will work together. The answer should be designed around reliability and schedule as well as energy price and emissions.

Cooling changes the energy and water equation

Higher computing density produces more heat. Cooling and environmental control can account for a relatively small share of electricity in an efficient hyperscale facility, or more than 30 percent in a less efficient enterprise data centre.

The choice of cooling system affects energy use, water demand, equipment design and site suitability. In water-stressed regions, developers must understand the source and quality of water, seasonal availability, discharge requirements and the trade-off between water use and electricity consumption.

This is especially relevant in the Gulf and parts of the Mediterranean, where high ambient temperatures can increase cooling requirements while water resources are already constrained. A site with inexpensive land and strong fibre connectivity may still be unsuitable if power and cooling cannot be secured on acceptable terms.

Resilience extends beyond backup generators

Data centres are designed around very high availability. Uninterruptible power supply systems and backup generation remain essential, but resilience now involves a larger set of dependencies.

Fuel supply for backup systems, transformer availability, spare equipment, cyber security, fibre diversity, access roads, emergency services and specialised maintenance all influence operating continuity. Rapid growth can also create competition for electrical equipment and skilled contractors.

These risks should be addressed during development rather than passed to the construction phase. The facility, utility connection and supporting services need one integrated critical path.

Governments and utilities need a clearer development framework

Data centres can support investment, digital services and economic diversification. They can also place concentrated pressure on grids, water systems and land planning. Public authorities need a method for assessing these projects consistently.

A credible framework should define connection policy, reinforcement costs, energy efficiency expectations, water requirements, environmental approvals and the treatment of on-site generation. It should also consider whether the project contributes to the local economy through digital services, employment, heat reuse, infrastructure upgrades or broader technology development.

Utilities need earlier visibility of the pipeline. If several large projects request connections independently, the system may respond project by project rather than planning a coordinated expansion.

The development team is changing

The traditional participants in a data centre project include the operator, developer, contractor and technology suppliers. The new scale of AI demand brings power developers, grid operators, storage providers, water specialists and public authorities into the project much earlier.

This changes consortium and governance requirements. Responsibilities for electricity delivery, energisation, cooling performance, carbon targets and infrastructure interfaces must be defined across several organisations. The project may also require phased capacity so that digital equipment and utility infrastructure can be commissioned in sequence.

AI has created a new class of digital demand, but the response is not purely digital. The projects that progress will be those that treat computing, power, cooling, water, connectivity and resilience as one infrastructure system from the beginning.

Further reading

Discuss this topic

Speak with MZA about how these dynamics apply to a specific programme or market.