Sovereignty Options: Delivering Control in Practice
Part 2 of a three-part series on sovereign AI

In the first article, we defined what sovereignty in AI means and why it matters. This second article moves from definition to delivery, examining the practical challenges boards must address when implementing sovereign AI systems.

As AI becomes embedded in core operations and critical infrastructure, sovereignty is no longer a technical consideration but a matter of governance, accountability and control. This article explores the realities of exposure, the spectrum of sovereignty options available, and the transition required to move from dependency to demonstrable control.

Designed for board-level audiences, it provides a structured framework to support informed decision-making, helping organisations clarify what they require from AI and how sovereignty can be achieved in practice.

The final article in this series will set out how these challenges can be addressed through models that are operationally viable, commercially sustainable and capable of delivering full sovereign control.

The question of what constitutes “sovereign AI” has been raised repeatedly in discussions with policymakers, investors, infrastructure providers and enterprise users, often framed in different ways but consistently returning to a common point of uncertainty: what does sovereignty actually mean in the context of AI systems, and how can it be demonstrated in practice rather than asserted in principle.

This paper sets out our sovereign view, defining sovereignty not as a label but as a test grounded in accountability, control and societal alignment within a jurisdictional boundary. It reflects a perspective formed through practical engagement with the challenges of deploying AI in environments where control, resilience and legitimacy are not optional characteristics, but operational requirements.

This is the first of a three-part series. In this paper, we define the conditions under which sovereignty can be said to exist. In the second, we examine how current delivery models address these conditions, and how a transition to full sovereignty can be achieved in practice. In the third, we set out how we deliver sovereign AI in practice, including how such delivery aligns with ESG requirements.

Eight SambaNova SN40 units have landed in Scotland, marking the launch of a sovereign AI infrastructure platform built for real-world deployment. Low power, air-cooled, and cloud-integrated, the platform is designed to scale rapidly as 150 systems are rolled out across the UK.

Argyll Data Development has entered into a strategic partnership with SambaNova to deliver the UK’s first renewable-powered sovereign AI inference cloud, hosted at the Killellan AI Growth Zone on Scotland’s Cowal Peninsula.

The 184-acre Killellan campus is being developed as a green digital hub, powered by a mix of on-site wind, wave, solar, biogas and waste-to-energy generation. A private-wire network and long-duration battery storage will supply firm, low-carbon power directly to the data centre, with waste heat captured and reused across the site for applications such as vertical farming, aquaculture and local heat loads.