Argyll Data Development has launched a UK sovereign AI inference cloud built with SambaNova. The general availability service makes SambaNova’s inference hardware available through a cloud hosted entirely within the UK.
The launch comes as governments and regulators pay closer attention to the energy use of artificial intelligence systems and data centres. Across Europe, policymakers have introduced measures to improve energy efficiency and ease pressure on electricity networks.
The new service is aimed at organisations that want AI inference delivered within UK regulatory jurisdiction. Argyll said the platform runs in existing UK data centres and does not require liquid cooling or specialist power infrastructure.
The arrangement marks SambaNova’s first general availability cloud offering in the UK. For British businesses, it provides access to the US company’s inference systems without placing workloads on infrastructure outside the country.
AI inference, the process of running trained models to generate outputs, has become a growing focus for businesses adopting generative AI tools. For many organisations, especially in regulated sectors, where that processing takes place and how much energy it uses are becoming key procurement questions.
Argyll said its platform supports open-source models including Minimax and can deliver speeds of up to 400 tokens per second. It positioned the service as suitable for large-scale agentic AI deployments, in which software systems complete tasks with limited human intervention.
Argyll argued that current AI infrastructure often forces customers to choose between performance, model quality and cost. It said its use of SambaNova’s Reconfigurable Dataflow Unit architecture was designed to avoid that trade-off while fitting into conventional UK data-centre environments.
Energy focus
The claim comes amid a wider debate over AI’s resource demands. The rapid expansion of large-model training and inference has raised concerns among policymakers, utilities and operators about how far existing power and cooling systems can support new computing loads.
Many advanced AI systems rely on hardware deployments that require high-density power supplies and liquid cooling. Argyll and SambaNova said their approach uses air-cooled equipment that can be installed in existing facilities without extensive redesign.
Peter Griffiths, Chairman, Argyll Data Development, said the service addresses a gap in the domestic market.
“Our platform gives UK organisations something they haven’t had before: premium AI inference running entirely within UK borders, at production scale, in existing data centres. It’s the foundation developers need to build agentic AI with confidence,” Griffiths said.
SambaNova, based in Silicon Valley, has been building AI chips, systems and cloud services for commercial and sovereign deployments. The partnership gives it a route into UK-hosted services as demand rises for national or regional control over data processing.
Rodrigo Liang, Chief Executive Officer, SambaNova, said many current inference deployments remain difficult for customers to adopt.
“High-performance inference usually requires liquid cooling, specialist power and purpose-built facilities, requirements that make sovereign deployment slow, expensive and out of reach for most,” Liang said.
He said the company’s underlying design was intended to remove those barriers.
“SambaNova’s RDU architecture was designed to remove those constraints: energy efficient, air-cooled and built to run inside existing data centres. Argyll has deployed that capability entirely within UK borders for the first time. It’s the model for how sovereign AI infrastructure should work,” Liang said.
Argyll is also developing a larger renewable-powered infrastructure project in Scotland. The company’s Killellan AI Growth Zone in Argyll spans 184 acres and is intended to combine wind, wave and solar energy with data-centre facilities.
The broader strategy reflects a growing push in the UK to align AI expansion with domestic infrastructure and energy policy. For developers and corporate users weighing compliance, location and electricity demand, the market for UK-based inference services is becoming more crowded and more politically significant.
