Industry views
15 Jul 2026

From cloud testbed to sovereign AI: what practical AI adoption teaches us about Britain’s digital infrastructure

Digital infrastructure is often discussed in abstract terms: cloud, compute, connectivity, data centres, platforms and networks. But for organisations trying to adopt AI, infrastructure is not abstract. It determines whether an idea can be tested in days or delayed for months; whether an SME can access advanced AI capability without major capital investment; whether a researcher can combine data and models safely; and whether a public body has enough confidence to move from experimentation to adoption.

Since October 2025, The Data Lab has been developing its AI testbed as part of the National AI Adoption Programme, funded by the Scottish Government. The purpose is simple: to provide Scottish organisations with a practical environment where they can develop, test and experience data and AI solutions without first having to procure specialist hardware, install complex software, or build their own AI engineering platform.

The testbed brings together the infrastructure required for modern applied AI: secure storage, IoT data management, managed AI and LLM services, model hosting, vector search, orchestration, notebooks, analytics, dashboards, identity controls and application deployment. It enables an organisation to move from a defined business or research problem to a working prototype that can be accessed through a secure cloud environment and tested directly by users.

That matters because most SMEs and many public-interest organisations do not fail to adopt AI for lack of ideas. They struggle because the enabling infrastructure is fragmented, expensive, technically complex or slow to access. A well-designed testbed changes the adoption pathway. It allows organisations to explore real use cases using real operational data, learn what works, and make better decisions before committing to full-scale implementation.

So far, The Data Lab’s AI testbed has supported 15 organisations across manufacturing, health, construction, education, creative industries, charities and AI product companies. It has enabled 12 pilots and proofs of concept, ranging from AI-assisted business workflows and knowledge retrieval to real-time monitoring, analytics dashboards and applied generative AI tools. The common thread is not the individual use case, but the infrastructure model: shared access to secure, scalable AI capability dramatically lowers the barrier to experimentation and helps organisations understand the value of their data.

The testbed is also supporting research and collaboration between academia and industry. Through a PhD placement with the National Records of Scotland, researchers can bring together data from different sources, apply AI and generative AI models, and develop applications that the host organisation can test directly by through a cloud platform. Without this kind of environment, similar work would typically require additional hardware, software installation, provisioning cycles and technical support before the research could reach practical users.

This practical delivery experience shows both the strength and the limitations of the UK’s current AI infrastructure landscape. Mature global cloud platforms are currently the fastest route to many AI capabilities. They provide developer tooling, managed services, scalability, and deployment options that enable rapid prototyping. They have allowed The Data Lab to support organisations quickly and cost-effectively, without forcing each SME or partner to build its own technical stack.

However, the same experience also shows why sovereign AI infrastructure matters. Several organisations, particularly public bodies and research partners, raise questions about where data is stored, who has jurisdiction over it, whether prompts and outputs are protected, how cross-border transfer risks are managed, and whether there is an assured environment suitable for sensitive or strategically important workloads. The issue is not that the global cloud cannot be used. The issue is that trust, assurance and jurisdiction concerns can slow or constrain adoption precisely where AI could deliver major public and economic value.

The next phase of digital infrastructure should therefore combine the best of both worlds: the speed and usability of modern cloud platforms with stronger sovereign, trusted and green AI capability. In practice, this means more than just data centres. It means sovereign AI inference platforms for public-sector and research use, national and regional GPU capacity for trusted workloads, “data factory” capability linking storage, governance and compute, accredited environments for SMEs and public bodies, and procurement routes that give start-ups, researchers and smaller organisations access to advanced compute. These are the capabilities that should be prioritised.

Scotland is well placed to contribute to that future. Its renewable energy capacity creates the potential for green AI infrastructure powered predominantly by clean electricity, including wind. With the right planning, grid access, heat reuse, water-resilient design, and a distributed compute strategy, Scotland could become a trusted place to develop, test and scale AI responsibly.

The lesson from The Data Lab’s AI testbed is clear: digital infrastructure is not a background utility. It is the foundation that determines whether organisations can innovate, whether public services can adopt AI safely, whether research can translate into usable tools, and whether SMEs can compete in an AI-enabled economy. The UK should invest in trusted, sovereign and sustainable AI infrastructure now to accelerate adoption while strengthening resilience, public confidence and long-term competitiveness.

 

Meet the Team 

Chris Hazell

Chris Hazell

Programme Manager - Cloud, Tech and Innovation, techUK

Josh Turpin

Josh Turpin

Programme Manager, Telecoms and Net Zero, techUK

 

 

Return to listing