There is a layer beneath the software, the models, and the skills that almost nobody is thinking about.

Large corporates have procurement teams, legal counsel, and technical architects to navigate AI infrastructure decisions. SMEs and scale-ups have none of that. They are signing contracts and committing workloads with no visibility of the long-term implications.

The layer that gets overlooked is the foundational one: compute access, energy costs, data residency, storage architecture, and vendor lock-in. Decisions made at pilot stage can lock an organisation into terms, jurisdiction, and pricing for three years or more, and they can be very hard to unpick at scale.

The risk looks different depending on who you are. For AI-native startups, early lock-in is a material risk, particularly as they scale. For traditional SMEs, the dominant concern is cost. Both need the same thing: a clearer view of the decisions they are making before they make them.

AI for Growth is developing a pragmatic procurement guide covering every layer of the AI infrastructure stack: what each layer is for, what to look for, what to avoid, and what questions to ask before you commit. It is built to empower small and emerging businesses to make future-ready decisions and protect themselves from extractive lock-ins and overseas provider risk.

The guide is being co-developed with experts across the storage, server, and software layers, run as a focused sprint and then released back to the market. The need is not limited to SMEs. The wider market lacks a resource like this.

The work covers the full infrastructure stack, from chips and supply chains, to data centres and neo-cloud providers, to data and models, applications, and services, so that any organisation can engage at the layer most relevant to it. Two initial focus areas anchor the work:

Infrastructure decisions shape what AI capability you can use, at what cost, with what reliability, and on whose terms. For a business without a procurement team or technical architects, those decisions are easy to get wrong and expensive to reverse.

AI for Growth's response is practical: tools and guidance that help small and emerging businesses ask the right questions and avoid commitments that lock them in. Government, suppliers, and consultancies all have a role in closing this gap. This pillar exists to help build the tools and provide the guidance to do it at scale.

It focuses on the foundational layer of AI adoption that is usually overlooked: compute access, energy costs, data residency, storage architecture, and vendor lock-in. The aim is to help SMEs and scale-ups treat infrastructure as the strategic decision it is, rather than an IT afterthought.

Decisions made early can lock an organisation into terms, jurisdiction, and pricing for years, and they are hard to unpick at scale. Large corporates have procurement and legal teams to manage this. Most SMEs and scale-ups do not, which is the gap this pillar is built to close.

Sovereignty is one of the pillar's two initial focus areas. The work is about building a clear, practical understanding of what sovereign AI means for the UK, rather than taking a position on it.

A pragmatic procurement guide covering every layer of the AI infrastructure stack: what to look for, what to avoid, and what questions to ask before committing. It is in development, co-created with experts across the stack and intended for release back to the wider market.