Artificial Intelligence is no longer the future. It is the present, and it’s driving increased demand for power. But the future isn’t AI-powered. It’s grid powered.
Rapid digitalisation and virtualisation means data centres play an increasingly central role in how we work and live. As data applications and bandwidth demand continue to grow, data centres are expanding at an unprecedented rate to support AI emergence and its production load requirements, but the power infrastructure is struggling to keep up. Utilities, bound by long planning cycles, strict regulations, and grid congestion, cannot expand capacity in the timescales data centres can be constructed. For AI data centre planners, this means the wait for grid connections is measured in years, not months.
Data centres have obtained power via the local power utility for decades, both for new sites and when expanding existing operations. But to keep pace with today’s surging demand, data centres must rethink how they work with infrastructure providers. Right now, data centres and utilities operate in separate silos: one makes requests, the other reacts. To break this cycle, the mindset must shift from “We need 1.5 GW tomorrow” to “What can we realistically and sustainably access next year, in two years, in six years?” Instead of rigid energy demands, a flexible strategy is needed that prioritises location intelligence and phased energy access.
The Challenge: Grid Congestion and Energy Constraints
- Surging Demand – AI demand is outpacing initial projections, and even as processing efficiency improves, energy consumption is unlikely to decrease. However, power grids take years to expand capacity making it almost impossible to keep pace while operating in the current siloed approach.
- Regulatory Barriers – Utilities are constrained by regulations and often cannot invest in additional capacity until demand is officially confirmed.
- Connection Queues – Lengthy grid connection periods are a reality worldwide as the decarbonisation of large parts of the economy requires many new connections. In the United Kingdom, for example, some connections to the transmission system will only be available in 2037, according to NESO’s latest Transmission Entry Capacity Register.
- Consenting and Permitting – The scale of AI demand means data centres require transmission connections that trigger much longer consenting and permitting processes, including, in some cases, environmental permitting and remediation work. These processes are beyond the control of regulators or utilities and therefore require reform at local and national government level if power is to be available to drive AI ambitions.
- Location Constraints – For power-dense data centres, location and proximity to energy sources, ideally renewable, is advantageous, but not always possible. For example, wind and solar energy in Texas is valuable but delivering that power to a data centre in Maryland is a challenge, logistically and financially. The location of fibre optic networks and the proximity of clients (for applications where data latency is critical) affects site selection.
These challenges are not going away but it is possible to accelerate connections to meet demand for AI if data centres and utilities meet in the middle. This requires a fundamental change in how data centres plan investments and how governments and utilities develop power infrastructure to support AI’s continuous growth.
A New Approach: Demand Flexibility and Smarter Siting
- Phased Connections – Instead of requesting and then waiting for full capacity, flexible energy access allows data centres to come online incrementally allowing rapid access to existing capacity while additional capacity is created.
- Load Flexibility – AI data centre demand is more variable than traditional data centre demand and AI workloads can shift dynamically across regions to optimise energy use and reduce stress on local grids. Utilities are working hard to accommodate such flexible connections, where demand varies over time but data centres may need to moderate ramp rates to meet utility requirements. This approach reduces the cost of infrastructure that is socialised across all network customers and can accelerate connections by avoiding peak network congestion periods.
- Moving from Reactive to Proactive Planning – To support the accelerating demand for energy by data centres, they and energy infrastructure should be part of integrated planning by local, regional and national/federal authorities.
- Energy Spatial Planning – Governments and utilities should proactively identify and incentivise strategic locations for all energy infrastructure and users, including data centres, encouraging them to consider sites where the grid is already strong, energy import/export is possible, and impact is minimised. For example, Great Britain’s system operator, NESO, is aligning its connection queue priorities with zonal energy requirements for demand and generation. A more direct intervention would be the use of nodal or zonal pricing. Considering the regional impact of data centres and the proportion of overall national demand being utilised is vital to limit network stability risk.
- Regional Energy Planning – Local, regional and national authorities can also signal where they are more open to data centre and energy infrastructure development. Combined with spatial energy plans this creates an efficient development, permitting and consenting platform. Victoria, South Australia, is developing renewable energy zones, pairing infrastructure investment with wind and solar projects to encourage data centres to connect in specific locations. This includes permitting and consenting of the data centres. Maryland’s clustering model, which emphasises pre-planned expansion and financing, has faced limited opposition. The UK’s AI Growth Zones seek to emulate this model.
- Incentivising Smart Siting – Tax credits, nodal or zonal energy pricing, and policy tools should direct development to areas with existing grid capacity. The issue is not just about cost or supply but about fostering cooperation between businesses and governments to invest in energy infrastructure in a way that benefits communities and industry alike.
- Policy and Industry Collaboration – Policymakers and industry leaders must take decisive steps to modernise energy planning for a digital-first economy. Utilities need policies that support proactive infrastructure investment.
- Public-Private Partnerships – Governments, utilities, and data centres must work together on strategic energy planning to avoid last-minute crises.
A Blueprint for the Future
AI-driven data centres are redefining energy demand but cannot operate within outdated infrastructure models. When data centres and utilities meet in the middle to find a flexible, location-conscious, and cooperative approach to accelerate deployments and prevent widespread grid failures, everyone wins. A step change is needed to ensure this approach becomes status quo.
