British AI company Locai Labs and UK sovereign cloud provider Civo have signed a Memorandum of Understanding to jointly develop Project Mercury, a family of pre-trained large language models trained and hosted entirely within the United Kingdom. The announcement, made on 14 April, positions the partnership as a domestic alternative to foreign-owned hyperscale AI infrastructure.
The Mercury Series, as described by the two companies, will span a range of model sizes: Edge Intelligence models between 0.8 and 30 billion parameters for low-latency local applications, and Frontier Power models at 256 billion parameters for complex generative tasks. Models will be available through Civo's UK-resident sovereign cloud or deployed on-premise within customer infrastructure. Training will run on 100 percent renewable energy.
The collaboration arrives shortly after the UK government confirmed its £500 million Sovereign AI Fund, part of a wider push to reduce reliance on US and Chinese model developers and hyperscalers. Unlike projects such as Stargate, which has dominated recent headlines around AI infrastructure, Mercury is pitched explicitly at British data residency and jurisdictional control.
James Drayson, Co-Founder and CEO of Locai Labs, said the partnership creates "a trusted, homegrown AI ecosystem that meets the highest standards of security, sustainability, and performance."
Mark Boost, Founder and CEO of Civo, framed the work as a demonstration that the UK can develop, train and host sovereign LLMs entirely domestically:
This partnership proves the UK can develop, train and host sovereign LLMs entirely on home soil, showing what two UK-founded companies can deliver for the security and trust of other UK-based enterprises.
A note on scope: the 14 April announcement is a Memorandum of Understanding, not a product launch. Training, benchmarking and availability timelines have not been published, and the 256 billion-parameter flagship remains a stated target rather than a shipping model. For UK buyers the practical question is when regulated workloads in finance, healthcare, defence and public sector can move to a Mercury-class model without the compliance overhead of routing through a US hyperscaler. The MoU sets the intent. The delivery plan has yet to follow.
