UK AI Data Center Trial Achieves 40% Power Reduction

A recent trial in London has shown that artificial intelligence-driven data centers can significantly reduce their electricity consumption, reaching up to 40% lower power usage during periods of high grid demand. This proof-of-concept marks a departure from the prevailing always-on data center model, which has contributed to increasing pressure on power grids globally.

The five-day test, conducted in December 2025 at a London facility, saw the participation of partners including Emerald AI, NVIDIA, National Grid, Nebius, and the Electric Power Research Institute. The trial involved over 200 simulated grid stress events. During these tests, the data center successfully lowered its energy draw by up to 40% while ensuring critical operations continued without disruption.

In practical scenarios, the facility dynamically responded to spikes in electricity demand. For example, it was able to reduce power consumption by 10% for as long as 10 consecutive hours—mimicking real-world situations such as halftime breaks during televised soccer matches, when grid demands surge. Notably, in one event, the center decreased its load by 30% within just 30 seconds.

"This trial proves that NVIDIA-powered infrastructure can act as a grid-aware asset, modulating demand in real-time to support stability," said Josh Paker, sustainability lead at NVIDIA. He added that making AI workloads adaptive to grid pressure not only aids energy reliability but can accelerate the deployment of AI infrastructure while minimizing the need for expensive grid upgrades.

The findings will inform the development of a planned 100-megawatt power-flexible AI facility that NVIDIA aims to operate in Virginia, USA. Data from the study is expected to be shared with the broader AI industry, energy regulators, and policymakers. The project participants hope that demonstrating the ability to adapt power consumption during peak periods could ease approval processes for new data centers seeking grid connections.

Steve Smith, president of National Grid Partners, noted the industry's interest in faster connection timelines, commenting: "We would love to get to a point where we can get customers on the network in two years, and this is part of that."

AI data centers are increasingly large consumers of energy due to the high-intensity computational workloads required for machine learning and automation. By leveraging advanced hardware, such as NVIDIA’s graphics processing units (GPUs), operators are aiming to add flexibility that matches workload intensity with grid availability.

As more industries and public services adopt AI tools, breakthroughs in infrastructure management—like those achieved in this UK trial—could play a vital role in balancing technical advancement with environmental and energy system sustainability.

Source: dataconomy.com