Colin Rees, Associate Director at IES, argues that the rejection of plans for an AI-focused data centre in Edinburgh should be seen as a warning to the wider sector: evidence, not ambition, will increasingly determine which projects move forward.

The decision to reject plans for an AI-focused data centre in west Edinburgh is not just a local planning story. It reflects a wider challenge now facing the UK data centre sector: how to bring forward proposals that clarify a project’s impacts in a planning environment that is asking harder questions about power, water and environmental impact.

Developers are not short of ambition; there is clear interest in investing in digital infrastructure across Scotland, and demand for capacity is being pulled forward by increasing AI workloads. But planning decisions are increasingly being shaped by whether project teams can evidence, in practical terms, how a facility will operate and what it will mean for local infrastructure, communities and the environment. As with any decision-making process, evidence is critical.

The tension is not new; what is changing is the level of scrutiny and the consequences when assumptions are left untested.

When evidence expectations are unclear, projects become risky

Environmental assessments are essential to the planning process, but in this relatively new AI world, there remains uncertainty around what evidence is expected, when it should be provided, and how potential impacts should be addressed. That uncertainty creates risk for everyone involved. Planning authorities can find themselves weighing competing claims without a consistent evidence base, while developers can face shifting requirements and prolonged engagement cycles as questions surface late. For developers under pressure to deliver capacity quickly, particularly those responding to surging AI-driven demand, these delays carry real commercial consequences. Time-to-power matters, and uncertainty in the evidence requirements makes timelines impossible to predict.

Recent experience elsewhere in the UK shows how quickly this can escalate. A proposed £1bn hyperscale data centre at Iver, Buckinghamshire, was approved on the basis of future power assumptions and has since faced a legal challenge. The details differ from project to project, but the underlying lesson is consistent: where critical operational impacts depend on assumptions rather than demonstrated analysis, approvals can become vulnerable.

The result is often mixed messages. Developers feel they are doing what is required, while planning authorities remain unconvinced that the evidence is complete or comparable. In that gap, objections harden and timelines stretch.

The sector is moving from commitments to operational proof

Data centres sit at the intersection of energy systems, cooling design and environmental performance. Yet too many proposals still lean on broad commitments rather than quantified scenarios. The industry needs to move beyond statements of intent and towards operational proof, presented in a way that planning stakeholders can interrogate.

Independent, evidence-based modelling is one of the clearest routes to gathering that proof. By showing how a data centre would operate in practice, teams can test different cooling, energy and design options and evidence how environmental impacts could be reduced. That changes the quality of the conversation. Instead of arguing in the abstract, developers can demonstrate what happens under different loads and system configurations, and they can quantify the impact of design choices on energy use, cooling demand and resource consumption.

This is also where whole-life performance matters. It’s not just about reporting; done properly, it is a way to connect physics-based understanding with how the asset behaves over time, supporting better decisions from design through operation and verification. In a sector under growing scrutiny, the ability to evidence whole-life performance – and to help close the performance gap between design intent and real operation – is moving from a nice-to-have to a requirement.

Cooling, water and predicted emissions are where planning debates intensify

Cooling strategy is often where sustainability concerns crystallise. The rapid growth of AI workloads has intensified this challenge dramatically. AI loads generate significantly higher heat densities than traditional IT loads, making cooling decisions even more critical – and the consequences of getting them wrong far more severe. Planning authorities are increasingly aware that generic cooling assumptions designed for conventional data centres may not be sufficient for AI-focused facilities. Depending on the approach, cooling can drive both water use and energy demand, which in turn shapes the environmental impacts a proposal may have. The key point is that these impacts are not necessarily fixed; they are design-dependent, and they can often be reduced materially when teams test alternatives early.

In real projects, modelling has already identified opportunities to cut reliance on water-intensive cooling, including one existing site where analysis indicated the potential to reduce water use by more than 90%. That kind of outcome is only visible when teams model operation in context, rather than treating cooling as something to be value-engineered later.

Energy efficiency is the second pillar, and it is becoming central to how proposals are judged. When facilities are subject to robust analysis from the outset, data centres can achieve industry-leading levels of energy efficiency, with power usage effectiveness values of around 1.2 or even lower. It is critical that this number represents annualised performance – an assessment of how the facility will operate across the entire year under varied loads and conditions.

However, the number itself is not the story. The story is what it represents: overhead energy that has been designed down through informed choices, supported by evidence, and framed in a way that can be assessed credibly.

Scotland’s opportunity depends on earlier, clearer engagement

AI-driven demand for data centres is not going anywhere. Scotland has an opportunity to attract investment and build digital infrastructure that supports economic growth and technological progress. But this will not be delivered on optimism alone. It will be delivered through proposals that are clearer, more robust and better aligned with environmental expectations from the outset.

Meeting sustainability goals does not have to come at the expense of progress. It does, however, require developers to engage earlier, be more transparent, and make full use of the technologies available to them. In practice, that means treating operational evidence as a front-end requirement, not a back-end defence. It means modelling impacts before positions harden. It means presenting options and mitigations as quantified scenarios rather than promises.

The Edinburgh decision should be read in that light, not as a rejection of data centres in principle, but as a signal that the sector’s planning playbook needs to evolve. The projects that succeed will be those that treat environmental performance as something to be proven – and then verified – not merely stated.