When Andy Burnham took over as Prime Minister in July, he created a never-before-seen role in his cabinet: AI Minister. Kanishka Narayan, who took up the post, called AI “likely the most significant technology in human history”. Building on the existing AI growth zones strategy, Narayan is developing a new plan for AI growth in the UK. Whatever he decides will significantly impact the UK’s future: from our energy resources to our global position as an AI innovator to our progress towards net zero.

What are AI growth zones? Examining the existing strategy

Before Burnham took over, the government already had an AI plan in place: the AI growth zones strategy. With the aim of multiplying the UK’s sovereign compute by twenty times by 2030, AI growth zones are areas where local authorities or companies plan to build 500MW data centres. These data centre construction plans can be submitted to the government and, if approved, become an official AI growth zone, with fast-tracked planning permission and access to power.

The problem? 500MW is way larger than the majority of data centres in the UK. Most operators won’t qualify as AI growth zones, or receive the benefits. Power is expensive in the UK; it doesn’t make sense to build huge hyperscale data centres with 500MW of capacity here. Currently, five data centres in the UK have qualified as AI growth zones, while the rest of the industry looked on, thinking: “AI growth zones are no use to us whatsoever”. It’s not only unhelpful for data centre operators, but also for the government, who aren’t seeing the AI growth they’d hoped.

How the AI strategy should be reshaped: a mandatory EPC for data centres

To boost AI growth without straining our power resources, we need an AI strategy that differentiates between different types of data centres. Hyperscalers, such as the 500MW data centres that qualify as AI growth zones, should be located in power-abundant places such as Scotland. Modular, AI-ready data centres, with capacity around 5MW, should be built in or near urban hubs, where power is more expensive. Locating these modular data centres in cities means they can easily connect to district heating networks and pass free, decarbonised heat back to the community. 

As well as differentiating data centres by their power requirements, we must also consider their environmental and community impact if we’re to stay on the path towards net zero. It’s impossible to sell a house without disclosing how energy efficient it is via an Energy Performance Certificate (EPC). Meanwhile, most data centre operators don’t disclose their metrics unless forced.

Beyond the EU taxonomy and ESG regulations, we need a mandated, data centre-specific way of reporting metrics: an EPC for data centres. The EPC works because it’s understood by industry figures and the public; it functions as a universal language for understanding efficiency in residential properties. If we’re to bridge the gap between data centre operators and the communities they operate in, we need a clearly defined way to communicate a data centre’s environmental and local impact.

The EPC for data centres in the UK should include the following metrics and provide a score to indicate overall efficiency, environmental benefits, and community benefits:

• Water Usage Effectiveness (WUE)

• Energy Reuse Factor (ERF)

• Power Usage Effectiveness (PUE)

• Renewable Energy Factor (REF)

• Cooling Efficiency Ratio (CER)

Then, the government’s AI strategy should prioritise grid access and planning permission for the data centres with higher scores.

Why sovereign compute is more essential than ever

When the US government added export controls to Anthropic back in June, AI models Mythos and Fable were disabled for all international users. This caused serious disruptions for UK businesses and government departments that relied on Anthropic’s AI models. Although the outage lasted just two weeks, it exposed how vulnerable the UK becomes when we rely on US-based companies for our AI technology.

As one of the people involved in crafting the government’s new AI strategy said to the Financial Times, “[We have to make sure] that 100 per cent of data centres aren’t owned by foreign companies.” Beyond the US, other locations for data centres would also put the UK’s data security at risk. While cheap power is readily available in the Middle East, data centres located there are especially vulnerable in the wake of the Iran conflict, with AWS facilities already targeted by drones.

Data centres in the UK, meanwhile, bring local tech jobs, data sovereignty, and, when facilities earn their social license to operate, positive benefits for nearby communities. We can use these modular AI-ready data centres for inference workloads that require low latency, as well as storing GDPR data and other sensitive information.

With the right growth strategy, the UK will thrive as an AI innovator

With the new AI strategy, the government has an opportunity to shape the UK’s future. Data centres are long-lived assets: the facilities that we build now will continue to impact our resource usage, environment and communities for decades to come. Prioritising sovereign, modular AI-ready data centres that demonstrate high efficiency as well as community and environmental benefits will ensure the UK thrives as an AI innovator. Curious to see a data centre like that in action? Check out our Urmston facility.