The UK’s first sovereign LLMs are here. Why this partnership matters

Locai Labs and Civo have partnered to develop the UK’s first pre-trained sovereign large language models. Developed, hosted, and operated within the UK, these models represent something the market has been missing for a long time. A serious attempt to build AI capability on foundations we actually control.

For all the recent momentum around AI in the UK, there is still a disconnect between ambition and reality. We talk about leadership, but much of what we build still relies on infrastructure and platforms owned elsewhere.

This partnership is about changing that.

From ambition to something real

There has been no shortage of intent.

The government’s AI Opportunities Action Plan and the creation of the Sovereign AI Unit point to a clear direction of travel. The recent Westminster debate on technology sovereignty showed that policymakers are starting to understand the risks of relying too heavily on a small number of global providers.

That shift matters. But it only becomes meaningful when it translates into real capability. If the UK wants to be a maker, not a taker, we need more than funding and strategy. We need infrastructure and models that are actually built and run here.

That is exactly what this collaboration is focused on delivering.

Introducing the Mercury Series

At the centre of this partnership is the Mercury Series. A new family of UK sovereign large language models designed to span everything from edge deployments through to large-scale cloud inference.

The goal is to build a complete spectrum of AI capabilities that organisations can deploy where it makes the most sense. On device, in private environments, or in the cloud. All developed, trained, and operated within the UK.

Jupiter is the first model in that series.

It is a large-scale, open source model designed with real-world use in mind. Transparent in how it is built, aligned to UK language and safety expectations, and flexible in how it can be deployed. Organisations can run it in the cloud or in fully controlled, air-gapped environments, depending on their requirements.

Together, the Mercury Series represents a shift towards practical, sovereign AI that organisations can actually use, adapt, and trust.

Why this actually matters

For years, the UK has been strong in AI research and early-stage innovation.

Where things tend to break down is at the point of production. That is where dependency becomes visible. Most organisations still rely on external infrastructure and models when they move beyond prototypes.

The Commons Library’s digital sovereignty briefing calls this out directly. Concentration of supply creates risk. Even when systems appear local, control often sits elsewhere.

This is where sovereign models change the equation.

When models are developed and operated within the UK, organisations gain clarity over governance and control. They know where their data is processed, which laws apply, and how systems are managed.

That is not just a compliance benefit. It changes how confidently organisations can adopt AI at scale.

True sovereign infrastructure

None of this works without the right infrastructure. Training and running models like Jupiter requires high-performance computing, fast storage, and low-latency networking. It also requires consistency. The ability to move from development to production without rebuilding everything along the way.

That is where Civo comes in. We provide the underlying platform that makes this possible. GPU-accelerated compute, Kubernetes native orchestration, and a cloud environment designed for modern workloads, all delivered within UK sovereign infrastructure.

The aim is not to make AI more complicated. It is to remove the friction that usually slows teams down once they try to scale.

Why now

The timing of this matters.

Government policy is moving in the right direction. Investment is increasing. The risks around dependency are now part of mainstream discussion. The recent Treasury announcement on AI investment makes it clear that the UK wants to compete. Competing in AI means building the full stack. Not just funding applications, but supporting the infrastructure and models that make those applications viable.

This partnership is one example of what that looks like in practice.

What comes next

This is not the end goal. It is a starting point. Building sovereign AI capability in the UK will take time and sustained effort. More infrastructure. More collaboration. More organisations willing to back UK-built technology.

But this is how it begins.

By combining Locai Labs’ model development expertise with Civo’s sovereign cloud infrastructure, we are starting to build something the UK has not had before. A credible, practical foundation for sovereign AI. If we are serious about being a maker, not a taker, this is exactly the direction we need to go in.

And now, we are moving.