TL;DR

Institutional capital is shifting toward UK private nuclear and fusion startups as the AI-driven datacentre buildout collides with volatile energy markets, according to market intelligence firm Tracxn. The firm has tracked roughly £283 million ($370 million) flowing into the sector, with £130 million ($170 million) of that arriving in 2024 alone, and is monitoring 83 UK nuclear startups clustered around Oxfordshire’s emerging “Nuclear Valley”.

The driver: AI demand meets energy anxiety

The government’s AI Opportunities Action Plan, published last year, has accelerated a domestic datacentre pipeline that needs always-on power at a scale UK renewables cannot reliably deliver. Combined with fuel-price volatility driven by the conflict in Iran, that is pushing investors toward sovereign baseload generation. Tokamak Energy, pursuing fusion, and Blue Energy, developing small modular reactors (SMRs), are among the firms drawing capital. Blue Energy is also supplying on-site power to a Crusoe AI datacentre in Port of Victoria, Texas — illustrating the “colocated reactor plus datacentre” model now taking hold in the US.

UK activity concentrates around Abingdon and Oxford, home to the UK Atomic Energy Authority, the MAST fusion experiment and Culham — designated the UK’s first AI Growth Zone. Edinburgh, Bristol and Glasgow are secondary hubs. Japanese industrial giants Hitachi and Toshiba have already made strategic buyouts of British atomic businesses, hinting at consolidation.

The awkward timing problem

Nuclear is a slow solution to an urgent problem. SMRs remain in development and fusion power is, on any honest assessment, at least a decade from commercial supply. A Centre for Net Zero report last year estimated it would be cheaper to power a 120 MW datacentre with renewables plus modest gas backup than with new nuclear — and AI compute demand is materialising in quarters, not decades. UK grid constraints and the Climate Change Committee’s pathway assumptions add further friction.

Looking forward

Investors are clearly betting that “baseload sovereignty” becomes a premium worth paying for over the 2030s, even if the maths does not work at today’s prices. For UK AI infrastructure policy, the near-term question is less about fusion and more about whether SMR approvals, grid connections and planning reform can keep pace with the datacentre pipeline already in the pipeline.