UK AI Strategy Needs Better Access to Computing Power

The UK is treating artificial intelligence as both a national security issue and a source of economic growth. At the UN General Assembly, Prime Minister Andy Burnham announced a new AI partnership between the UK and US to protect critical infrastructure, including sea lanes and airspace. The announcement raises a practical question for British AI companies: where will they get the computing power their work requires?

Companies need access to GPUs to train and run AI models, but large cloud providers and AI firms are securing capacity years in advance. Smaller companies can struggle to obtain computing power on terms that fit specific projects. An Oxford University study found that specialized AI data centers operate in just 32 countries, and US and Chinese firms run more than 90% of them. American technology companies operate 87 major AI hubs worldwide, compared with six run by European companies.

Demand is high: the UK government's Sovereign AI Fund received more than 200 applications seeking over 67 million GPU hours in three months. At the same time, an analysis of around 23,000 cloud clusters found average GPU utilization of 5%, as some organizations hold capacity that is not being used.

The article argues that pooling underused GPU capacity from data centers, businesses and other providers could give British companies another source of computing power under UK jurisdiction. The government has committed over £2 billion to public compute, including plans to expand the AI Research Resource twentyfold by 2030. The article says demand from startups already exceeds what is available, while much of that investment is going to a small number of national supercomputers.