Why Scotland is putting the brakes on AI data centres
Subscribe to our free newsletter today to keep up to date with the latest construction and civil engineering news.
Scotland is moving to slow the rapid growth of AI data centres as concerns rise over electricity demand, water use, emissions and pressure on local communities.
MSPs voted on Sept. 16 to support a pause of up to 12 months on planning decisions for new AI data centres while national guidance for large facilities is developed. The vote came as more than 20 large data centres have been proposed across Scotland.
At the same time, the Scottish Government has introduced stricter environmental checks for large projects. From Sept. 17, new data centres with a power capacity of more than 50 MW require an Environmental Impact Assessment. Smaller developments can also face an assessment if local conditions warrant one.
The two measures point to a wider shift in Scotland’s approach to AI infrastructure.
The country remains interested in the investment and jobs linked to data centres. But policymakers are now asking whether the planning system, electricity network and environmental rules can support the scale and pace of proposed development.
AI data centres are becoming an energy planning issue
AI has increased the amount of computing power required by technology companies. That has made large data centres an increasingly important part of digital infrastructure.
It has also increased their physical demands.
The Scottish Government says AI development raises national questions around energy, water use and hardware needs. Its environmental guidance also highlights possible effects from construction, greenhouse gas emissions, biodiversity, water resources, noise, landscape changes and air pollution from backup generators.
Electricity is likely to receive particular attention.
In a response published in May, the Scottish Government said it had not completed a dedicated assessment of total future electricity demand from AI and data centre development in Scotland.
It did, however, point to the National Energy System Operator’s 2025 Future Energy Scenarios. That work indicated that up to one-fifth of Great Britain’s future data centre electricity demand could be located in Scotland.
That creates an important issue for developers.
Scotland has significant renewable energy resources, which can make it an attractive location for companies seeking lower-carbon computing capacity. Yet high levels of renewable generation do not automatically mean large new loads can be connected where and when developers want them.
Grid capacity, network upgrades and connection times can still limit development.
The Scottish Government had already started increasing its oversight before the September vote. Since Aug. 17, local planning authorities have been required to notify ministers within seven days when they receive an application for a data centre with more than 50 MW of power capacity.
The government said the measure would help it monitor the development pipeline and understand its effect on wider infrastructure.
The planning debate is therefore moving beyond individual sites. Scotland is starting to consider data centres as part of a national infrastructure issue.
Environmental scrutiny is catching up with AI investment
The new rules also show how the environmental debate around data centres is changing.
Data centres have long been assessed through factors such as land use, local planning and energy efficiency. AI is adding scale to those concerns.
Larger computing facilities can require substantial electricity supplies and cooling systems. The Scottish Government has said closed-loop cooling can reduce water demand and renewable power can improve energy performance.
It has also warned that data centres using fossil fuels could produce emissions that are not accounted for in Scotland’s existing Climate Change Plan.
For developers, that raises the level of evidence likely to be required at the planning stage.
A project may no longer be judged mainly on whether a suitable site is available. Applicants may need to provide a stronger account of where electricity will come from, how cooling will work, what backup power will be used and how development will affect nearby communities.
This does not amount to a permanent block on data centre construction.
The Sept. 16 vote called for planning decisions to be delayed for up to one year while national guidance is prepared. Separately, the mandatory environmental assessment rule applies to projects above the 50 MW threshold.
The distinction matters because Scotland appears to be developing a framework for future growth rather than closing the market.
Scotland may offer an early lesson for the wider AI buildout
The debate in Scotland reflects a problem likely to appear in other markets.
AI development is often discussed in terms of chips, software and investment. Data centres show that growth also depends on electricity networks, land, cooling systems, construction capacity and planning approval.
Those systems can move at a different pace from investment in AI.
Technology companies can order computing hardware and raise capital relatively quickly. New transmission infrastructure, substations and generation projects can take much longer to plan and build.
That creates a risk that proposed data centre capacity grows faster than the infrastructure needed to support it.
Scotland’s response suggests governments may intervene earlier when those pressures become more visible.
For data centre operators, the commercial case for a location may therefore depend on more than electricity prices or access to renewable generation. Developers may also need to consider grid availability, environmental assessment requirements and the ability of local infrastructure to support large new loads.
Scotland still has reasons to attract AI infrastructure. Data centres can support investment, digital services and employment while demand for computing capacity continues to rise.
But the latest planning measures show that those benefits are being weighed against increasingly visible physical limits.
The next phase of AI infrastructure growth may depend as much on power systems and planning policy as it does on advances in computing.
Source
