Author: Izzy Woolgar, Director of External Affairs, Centre for Net Zero

The global race to dominate artificial intelligence (AI) will no longer be won by countries with the most advanced chips, but by those with access to large amounts of power, and the ability to deploy it quickly. In this race, power is the lane that matters most - and the UK risks being left at the starting line.

The physical constraint threatening to throttle big tech is no longer silicon, but electricity. Whilst the global AI bubble might be overhyped, the gap between demand and supply is not going away in the short to medium-term. As OpenAI’s Sam Altman recently posted on X: “we believe the risk to OpenAI of not having enough computing power is more significant and more likely than the risk of having too much".

If this trend were to reverse in future and we end up building ahead of need, we can be confident that new use cases will emerge. The tech sector’s appetite for energy has always expanded to fill the infrastructure available, and Netflix is the prime example: rapidly pivoting away from its DVD-by-mail service thanks to the increasing availability and capacity of high-speed internet.

In many markets, meeting new sources of demand is complicated by the energy transition. Scaling AI while successfully decarbonising power systems is one of the biggest industry challenges we face - and the reality is here on our doorstep. Tech giants are urgently searching for quick ways to procure ideally clean large amounts of electricity, in a context of painfully slow grid expansion - creating acute tension between the UK’s desire to build and the limits of our current energy system.

The transition’s growing pains

The pivot to renewables was meant to deliver cheaper power. But whilst renewable generation costs have plummeted, retail and industrial electricity prices remain high. Building out the network, subsidising profit-hungry investors and handling increasing variability and system imbalances comes at a cost. The result is a messy, retrofit-style transition that is neither fast nor efficient.

This leaves Big Tech with two acute challenges. The first is long delays in securing grid connections, and the backlog appears to be growing at speed. In just six months, the National Energy System Operator has seen a 470% increase in demand connection applications at a transmission level (see graph below) - a significant share of which have come from data centre operators. The second is high electricity prices. The UK’s electricity prices are among the highest globally, roughly four times those in the United States and well above the median of comparable countries.

Enter decentralisation

We are now at the point where bypassing the central grid makes economic sense. Microgrids and private wires that connect directly to new generation sources offer businesses a route to power at speed, independent of central system bottlenecks.

This decentralised model aligns with broader market shifts. In the UK, as investment becomes guided as much by planning and regulation as by market prices, grid defection allows investors to grow capacity alongside local demand while avoiding the legacy costs of national networks.

Growing frustration with delays is already driving data centre operators towards direct gas connections. Yet there is a cleaner alternative. Recent modelling from Centre for Net Zero shows that a hybrid microgrid - powered by 150 MW of offshore wind, 94 MW of solar, 100 MW of gas and 156 MW of battery storage - could meet the needs of a 120 MW data centre at an average cost of £71/MWh, compared with grid prices exceeding £250/MWh.

Crucially, this approach carries social legitimacy compared to other options that could offer speed to power, such as gas or nuclear power. Public sentiment remains firmly in favour of green digital growth: 76% of Europeans support data centres powered by renewables, and 72% favour rules requiring new facilities to build new clean generation alongside them. In an era where social consent for the transition is fraying, this matters.

Towards a new approach

Outdated assumptions must now be challenged. Data centres do not require firm grid connections for the entirety of their energy load, nor do they operate at maximum capacity continuously. In reality, whilst good data is lacking, utilisation rates are estimated to sit between 50% and 80%. This reality supports a more flexible approach as the default. It also means that rather than sourcing significantly sized grid connections, based on a data centre’s full load, partial grid connections could instead be utilised and supplemented with dedicated microgrids. This provides resilience, and allows data centres to get operating quickly.

There are early signs that the tides could be turning towards decentralisation. If we want to better understand what might lie ahead for our tech and energy sector, we need only glance across the pond. Last month, one of the first-known deals in the world was made to deploy battery storage at scale to accelerate speed to power for a US data centre in the Pacific Northwest. Planned to be operational in 2026, a 31MW battery storage system (BESS) enables the facility to come online and scale operations years earlier than would have been possible with traditional utility upgrades. To date, few BESS data centre deals have been announced. With more reportedly in the pipeline, it feels as though a new, effective model is emerging - and regulators on both sides of the Atlantic are taking notice.

Despite US government rhetoric around solar, wind and batteries, earlier this month, US Energy Secretary Chris Wright wrote to the US regulator asking them to consider rule changes to accelerate data centre connections where operators commit to flexibility or co-locate with dispatchable generation. Meanwhile, the UK’s Department for Science, Innovation & Technology has recently proposed that “AI Growth Zones” provide pricing mechanisms to reward data centre operators who provide flexibility and help manage local grid constraints.

The race ahead

So who will win the AI race? China’s growing energy advantage could offset its relative weakness in chips, potentially reshaping global leadership and closing in on the US. But the outcome is not pre-ordained.

For the UK to compete, speed to power must become a national priority. Its strategy should focus on levers including: incentivising data centres to share operational data to unlock flexible energy consumption, fast-tracking decentralised power solutions, and leveraging its world-class offshore wind resource to deliver clean, affordable electricity at scale.

The economic prize is substantial. Adopting these policies would strengthen industrial competitiveness, stimulate local economies, and lower consumer bills. If the UK can lift its capacity above recent trends, there’s a reported £44 billion in value to the economy over the next decade on the table. And as geopolitical fragmentation accelerates, expanding our sovereign compute capacity is as much about national security as it is growth and innovation.