Rising Power Demand from AI Data Centers

Developers of artificial‑intelligence (AI) data centers are queuing to connect to power grids across Europe, creating a bottleneck that threatens the continent’s ambition to host a share of the massive compute spending by AI labs. Grid operators highlight that the primary constraint is not the availability of energy but the ability to transport it to the sites where it is needed. In the United Kingdom, National Grid reports that data‑center applications representing more than 30 gigawatts of power demand are awaiting connection—an amount comparable to two‑thirds of Britain’s peak demand. The surge has caused some projects to be cancelled because of insufficient grid access.

Current Infrastructure Limits

Existing transmission networks were designed for traditional loads and are now facing a new wave of electricity consumption from AI workloads, which are both large and intermittent. The geographic mismatch between renewable generation—largely in Scotland and northern England—and consumption centers in the southern, more densely populated regions further complicates matters. Building new high‑voltage lines is a long‑term solution that can take seven to 14 years, given planning, legal, supply‑chain and labor challenges.

Grid‑Enhancing Technologies as a Stopgap

To squeeze more capacity from the current grid, operators are experimenting with several technologies. Dynamic line rating (DLR) uses sensors to adjust the amount of electricity a line can carry based on real‑time weather conditions, allowing higher throughput when lines are cooler and windier. Neara’s managing director notes that about three‑quarters of the UK network could transport more energy if DLR were widely applied. An EU study suggests that such “grid‑enhancing technologies” could boost overall capacity by as much as 40 percent in theory.

National Grid plans to deploy DLR on many of its busiest circuits within the next two years, though it has so far installed the system on only 275 kilometers of line. Operators also explore line‑bypass methods that divert power around congested sections and demand‑flexibility programs that enable data centers to shift consumption in response to grid conditions. Because AI workloads can be more flexible than traditional compute, they may be able to reduce usage or draw from onsite batteries during peak strain.

Challenges and Regulatory Outlook

Despite the promise of these technologies, there are limitations. During heat waves, when data centers need maximum cooling, the grid’s capacity may actually decrease, making it unsafe to run additional power. Moreover, current regulations prevent National Grid from counting data‑center flexibility in connection planning, limiting the impact of these measures.

Ofgem is preparing reforms to prioritize serious proposals and penalize operators that fail to expand capacity or meet connection deadlines. The regulator emphasizes the need to connect data centers quickly to maintain competitive advantage.

Future Infrastructure Needs

National Grid estimates that, over the past five years, it has expanded network capacity by 16 gigawatts through a mix of new conductors and grid‑enhancing technologies. However, officials acknowledge that to double the amount of energy flowing over the network in the next five years, significant new overhead lines will be required. The combination of short‑term technological upgrades and long‑term infrastructure investment is seen as essential to accommodate the growing AI compute demand while preserving grid reliability.

Used: News Factory APP - news discovery and automation - ChatGPT for Business