BESS & Grid Storage Developed 2025 · C13 4 min

Battery Energy Storage Systems for AI Data Centers in the UK

BESS for AI data centers in the UK addresses two problems at once: keeping compute-critical facilities online and making better use of a strained electricity grid. Even brief power disruptions can cause costly downtime and threaten AI operations, and battery energy storage provides instantaneous backup, smooths fluctuations and manages peak loads. This case study examines the business case for co-locating large-scale storage with AI data centers, and the socio-techno-economic, technology and policy factors that decide whether it works.

A constrained grid meets surging demand

The UK grid is increasingly congested. In 2024, about 10 percent of planned wind output was curtailed, rising to as much as 20 percent in Scotland, as more than 40 GW of renewable capacity has been added since 2000. Total nameplate capacity around the British Isles now reaches roughly 105 GW against peak demand of only about 50 GW, and the National Grid has warned that annual curtailment costs could hit 2.5 billion pounds by 2030. At the same time, AI compute is driving aggressive growth in data center demand. Both developments run into the same wall: grid connections are limited, oversubscribed, and slow and expensive to upgrade. The UK also has some of the highest wholesale electricity prices in the world, which weighs heavily on any data center business case. Co-locating storage above 10 MW with AI data centers is framed as a way to make scarce grid infrastructure work harder.

The co-location business case

Several synergies underpin the argument. A shared grid connection can serve both load and energy trading, so each megawatt of capacity earns maximum return rather than sitting idle. Layering storage onto data center schemes reuses sunk investment in land, planning and grid connections, lifting asset utilisation and internal rate of return without new consents. Load profiles complement each other: data center cooling demand is higher in summer, while storage trades most actively in winter when system stress and price volatility peak, so import and export capacity stays busy year-round without conflict. The connection can also host other power-intensive activities such as high-performance computing or GPU farms. Locating near curtailable renewables via private-wire links lets the data center buy low-cost zero-carbon power while storage captures arbitrage on excess generation, reducing exposure to volatile wholesale prices and use-of-system charges.

Technology and revenue

Data centers today rely on double-conversion online uninterruptible power supplies (UPS) and diesel generators. A UPS holds critical IT load for a short window, typically 5 to 15 minutes, long enough for generators to take over. Storage sits alongside the UPS on a different timescale: where the UPS covers milliseconds to minutes, the battery system provides power over hours and can eventually reduce reliance on diesel. Among the five broad storage categories, electrochemical systems fit the requirement best, offering high efficiency, scalability, modularity and compatibility with renewables, along with the long cycle life needed for cost-effective operation. Storage for AI facilities must deliver high power density, respond rapidly during outages, integrate cleanly with existing infrastructure, and scale as demand evolves. Combining contracted data center cash flows with merchant trading and grid-services revenue from storage diversifies income, lowers concentration risk, reduces the weighted average cost of capital and makes projects more bankable.

What it means for the UK market

The UK is a world leader in renewable penetration, with over half of generation now met by renewables and a large share from wind, but that same profile creates instability and curtailment because variable generation lacks the inertia of centralised plants. Storage helps stabilise frequency and voltage while unlocking trading opportunities, and pairing it with data centers turns two new asset classes into one investment package attractive to long-term infrastructure and energy-transition funds. The core insight is that grid constraint is the binding problem, and co-location converts curtailment cost into value-added output. For developers and investors, the model improves returns on scarce connections, diversifies revenue, and aligns digital and energy infrastructure that increasingly draw on the same pool of capital.

Key Takeaways

The UK grid is constrained, with about 10 percent of 2024 wind output curtailed and up to 20 percent in Scotland.

Curtailment costs could reach 2.5 billion pounds a year by 2030 as capacity of roughly 105 GW meets peak demand near 50 GW.

Co-locating BESS above 10 MW with AI data centers makes scarce grid connections serve both load and trading.

Reusing sunk investment in land, planning and connections lifts asset utilisation and internal rate of return.

Data center and storage load profiles are counter-correlated across seasons, keeping shared capacity fully used.

Electrochemical storage best fits the requirement for high power density, scalability and long cycle life.

Blending contracted data center revenue with merchant storage income lowers the cost of capital and improves bankability.