Agentic AI Risks Supercharging UK Data Centre Power Demands

Data centre power usage is projected to increase by 17 per cent in 2025 alone, doubling by 2030, according to the IEA. Agentic AI systems, which are autonomous and perform complex tasks, could demand significantly more power per query; one study found they used up to 136.5 times more energy than a basic chatbot session. A scenario where daily agentic AI requests match current Google searches could create nearly 200 gigawatts of new power demand, about half the current US power grid load.

The article highlights that in the US, data centre developers are commissioning their own power plants, often gas-fired, to secure electricity supply amidst rising demand. This trend, if mirrored in the UK, could undermine efforts to decarbonise the electricity supply and create new dependencies on fossil fuels, posing a challenge to the UK's energy security and climate commitments.

Businesses deploying agentic AI at scale will face increasing energy costs and procurement challenges. The article notes that data centres' rising energy demands have already pushed electricity prices higher in parts of the US. Companies developing AI solutions, such as Meta with its 'Muse' agent, are driving towards mass consumer adoption, which could further amplify energy requirements.

While not directly stated, the article implies potential indirect impacts on consumers. Rising electricity demand from AI data centres could contribute to higher energy prices, as has been observed in parts of the US. Increased reliance on fossil fuels to power these centres could also affect air quality and contribute to climate change, with broader societal costs.