1. AI’s future hinges on physical infrastructure

AI depends on computer servers, data centres and grid connectivity. Running these physical systems requires huge amounts of electricity to power them, while data centres can also consume large volumes of water for cooling.

With AI use increasing, countries must treat compute capacity – the computing power needed to train and run AI systems – as strategic infrastructure, not a background technicality. AI should be considered a physical industrial system, and its success depends on energy availability, land use and countries' capacity to build and power data centres at scale.

2. Energy constraints could become a bottleneck to AI growth

Tom Cope, partner in Deloitte’s infrastructure and capital projects team, argues that AI demand has arrived at “the most imperfect time.” Electricity networks are already feeling the strain, with supply chain pressures on materials to develop the grids and rising energy prices. And with the cost of electricity varying significantly across regions, there are limits on where data centres can be built successfully.

This creates tension on a global scale. Nations with cheaper, more secure energy will attract compute investment, enhancing their AI competitiveness.

3. Data centre placement causes local friction despite national benefits

Inference data centres, which enable rapid responses for GenAI tools, provide speed by being built close to cities. But communities don’t always want large industrial buildings on their doorstep. As Caroline Brown, partner in AI for infrastructure and sustainability, puts it, “Everybody wants all of the benefits of AI. Nobody wants to look at it at the end of their garden.”

Land use, water consumption for cooling and fears about electricity competition all raise local concerns. Yet proximity to these data centres may also have benefits, as they can help balance grid demand while waste heat can warm nearby homes and even swimming pools. The challenge is aligning local impacts with national ambitions.

4. AI could significantly accelerate sustainability and energy efficiency

There are valid concerns about AI’s energy use. But Caroline highlights research that shows advancements in AI could reduce global carbon emissions by five billion tonnes over 15 years. That’s five times more than it generates, so a benefit ratio of five-to-one.

The National Energy System Operator is already using AI to help balance the flow of electricity between homes and other uses, such as businesses, EVs and data centres. AI also has the potential to accelerate the development of low carbon materials, decarbonise industrial processes and improve forecasting across energy systems.

5. Energy access, compute capacity and strategy will become competitive advantages

As demand increases and less compute becomes readily available, organisations will need to decide whether higher spec, higher cost data centres – likely with more capacity – are a better strategic fit for their plans, or if they want to be “competing in a much more crowded, cheaper end of the computer market” for simpler use cases.

The businesses that ensure compute and physical AI needs are central to their commercial strategies will have an advantage over firms still figuring out how to navigate an evolving market.

Contact
Mark Heads
Podcast and Brand Content Execution Lead
mcheads@deloitte.co.uk