Argyll, SambaNova open UK sovereign AI inference cloud in Scotland

Argyll Data Development and SambaNova have opened a UK sovereign AI inference cloud at the Killellan AI Growth Zone, a 184-acre digital campus on the Cowal Peninsula in Scotland. The facility is live and runs SambaNova’s Reconfigurable Data Unit (RDU) hardware with SambaManaged software.

All data, models and operational control are retained inside the UK. Operations are carried out by UK-resident staff, and the site is powered by 100% renewable Scottish energy.

The service is designed for inference workloads-applying trained models to live data-rather than for model training. SambaNova’s RDU architecture is built to operate within a power envelope of about 10 kilowatts per rack, which the partners say reduces cooling and energy requirements compared with many training-focused systems.

Argyll configured the cloud as a disaggregated platform so compute, storage and networking can sit in multiple UK locations while presenting a single unified inference layer. Argyll says that arrangement increases resilience and allows organisations in regulated or security-sensitive sectors to keep infrastructure and data under local control.

The partners expect early customers to come from life sciences, medicine and banking, where low latency and data residency are common requirements. Other anticipated use cases include logistics and customs management, education and agriculture. The cloud will host open-source models such as Minimax; Argyll reports Minimax can run at speeds up to 400 tokens per second in the UK-resident environment.

Jude Sheeran, managing director for EMEA at SambaNova, said: “Many organizations default to GPU infrastructure without fully accounting for long-term cost, energy and operational complexity. Our work with Argyll provides an alternative, enabling high-performance AI inference that is more efficient, deployable and aligned with sovereignty requirements.”

Peter Griffiths, chairman at Argyll Data Development, stated: “Sovereignty in AI is not a label you can apply to a contract or a colocation agreement. It is a condition that has to be demonstrated — who is accountable, where the infrastructure sits, who controls the intelligence layer, and whether all of that aligns with the expectations of the society being served. Our platform satisfies those conditions. We are building the standard that others should be measured against.”