From burden to backbone: The data centre’s new role in the power system

The traditional model treats the data centre as a passive load: connecting to the grid, relying on utility supply and only using backup systems when needed. That model worked when data centres were smaller, predictable facilities with steady megawatt-scale demand. But it doesn’t hold at today’s scale.

The rise of AI data centres has transformed megawatt facilities into gigawatt campuses, changing how much power is required and how it is drawn from the grid. A single large connection can concentrate demand at a single node, compressing decades of incremental growth into one point on the network. 

For grid operators and data centre developers/owners, AI data centres have created new technical challenges. Stability now depends on reactive power, inertia and grid strength, and on the ability to maintain voltage and frequency under sudden changes in demand. The core design brief: do not destabilise the system.

AI training loads create large, cyclical swings at gigawatt scale while usage-driven AI inference requests can cause frequent and difficult-to-predict spikes in power consumption. Data centres are now dynamic power assets with the potential to stress or destabilise the grid if not designed at a system level.

Across Europe and North America, data centre developers are increasingly bringing their own power because the grid cannot connect them quickly enough. Connection lead times extend years beyond project timelines, making traditional reliance on utility supply commercially unviable.

Behind-the-meter power generation, often referred to as “bring your own power” (BYOP) or bridging power, has emerged as a direct response to these delays. On-site energy systems using gas turbines or engines, combined with battery storage, renewables or long-duration supply, can secure power, reduce connection delays and help campuses come online ahead of full grid connection.

However, bringing your own power comes with its own technical and commercial challenges. Turbines and major electrical equipment take time to procure and install, while operating in island mode shifts responsibility for reliability, resilience and power quality to the developer. A sophisticated microgrid controller must actively balance generation, storage and load, manage distribution across the campus and respond to changing demand while maintaining uptime and power quality.

The power architecture itself is a more complex, integrated system. Behind-the-meter energy centres may combine gas generation, battery storage, renewables and longer-duration supply. Batteries add flexibility by buffering grid fluctuations and fast-changing AI loads, reducing stress on the grid, smoothing demand peaks and improving generator efficiency.

For example, Microsoft’s purchase of the Three Mile Island nuclear power plant in Pennsylvania shows how data centre operators are seeking to move beyond traditional grid supply and towards dedicated, long-duration energy arrangements.

Power now frequently shapes data centre site selection, permitting and delivery strategy. A viable site requires more than land, fibre and customer demand. It needs a clear energy pathway, including fuel supply, emissions considerations, grid connection requirements and approvals for generation infrastructure, alongside how the project responds to local community concerns about energy use, cost and environmental impact. In many regions, communities are questioning how large new loads interact with existing energy systems and who benefits from the investment. The power solution is as significant as the data centre itself.

Assets built to overcome grid constraints can also be designed to support the grid itself. Bridging power assets can support the grid when on-site generation, storage and microgrid control systems are configured to operate as part of the wider power network, extending the data centre’s role from a consumer of electricity to an active participant in system stability.

Battery storage, flexible generation and control systems can provide services that grids increasingly need, including demand shaping, frequency response, voltage support and recovery after disruptions. Over time, these assets can move from temporary power solutions to being part of the grid’s operating infrastructure.

Data centres are already changing their relationship with the power system. While they remain large electricity users, they can also support the grid by adjusting demand, using storage or dispatching on-site generation when needed. For example, Google has agreed to make some of its data centre power demand flexible by temporarily reducing or shifting certain computing tasks during peak demand periods. This will help utilities manage pressure on the grid while supporting faster connection of new data centres.

Designing for the future requires planning the energy system from the start. The assets and fuels behind the meter need to follow a long-term pathway, rather than simply solve a short-term constraint. Some projects may reuse existing power infrastructure from former mining or industrial sites, while others may adopt flexible fuel strategies such as biomethane, hydrogen or ammonia, supported by emissions controls or hydrogen-ready equipment.

In Finland, data centre waste heat is reused for district heating, while in the UK, data centres are rolling out programs that feed free heat to swimming pools. When designed as integrated energy assets, data centres can help strengthen local resilience and become a constructive part of the power system.

For data centre developers, power is now a core development decision. Each project needs an early energy strategy that aligns grid access, on-site generation, storage, fuel supply and emissions requirements with the site’s delivery plan.

Bridging power should be designed with the future grid in mind. When built for flexibility, stability and lower-carbon fuels, it can evolve into long-term infrastructure that supports both the campus and the wider system.

Developers should also expect tougher grid requirements over time. As data centres scale from megawatt facilities to gigawatt campuses, grid operators are increasingly likely to call on developers to do more to support reliability and system stability. Large new loads can place pressure on shared grid infrastructure, which is driving closer regulator and community scrutiny of how data centres are funded, integrated and aligned with broader energy transitions. Planning for these shifting expectations early reduces risk and avoids costly redesign.

Developers and grid operators should:

• Treat the power solution as a core design problem shaping data centre site selection, permitting and delivery.
• Design bridging power as a future grid asset, incorporating storage, stability systems and fuel-flexible generation from the start.
• Plan for evolving grid requirements, including regulatory and community obligations around stability, cost, reactive support and demand flexibility.

There is a clear choice. Data centres can remain large, inflexible loads that add pressure to constrained networks, or they can be designed as integrated energy assets that help manage demand, strengthen local resilience and support the grid. In the new power system, the data centres that succeed will be those designed to strengthen the system they depend on from the outset.

Talk to us about how an integrated energy strategy can secure power earlier and position your data centre as a long-term asset to the grid.