It is easy to talk about AI as though it lived entirely in software. In practice, AI Growth Zones are fundamentally about electricity. Data centres are large, sustained electrical loads, and the scale involved is significant: South Wales alone is expected to draw more than 1GW of power in the early 2030s. Decisions about where that power comes from, how the grid is reinforced, and who gets connected first will shape the Welsh economy for years.
That is why the third test in this campaign's framework is about grid reinforcement and renewables. If planning treats power purely as an input for large operators, smaller firms can find themselves at the back of a very long queue. And a firm that cannot get an affordable, timely grid connection cannot easily expand, electrify its processes, or adopt the AI-enabled equipment that raises productivity.
AI is a power story
Adopting AI is not only about buying software. For many Welsh SMEs, especially in manufacturing, food and drink, agriculture and logistics, it is bound up with electrification more broadly: automated equipment, sensors and monitoring, electric vehicles and machinery, refrigeration, and the processing power to run modern tools. All of that needs reliable, affordable electricity and, often, an upgraded connection.
So when a Growth Zone concentrates enormous demand in one place, it is not a neutral event for the firms nearby. It can raise the stakes on an already constrained grid. If capacity and reinforcement are planned well, major AI infrastructure can be the reason the grid finally gets upgraded in a region. If they are planned badly, it can crowd out the very SMEs the zone is supposed to help.
The connection queue problem
Across the UK, one of the most persistent barriers to investment has been the time it takes to get a grid connection. Businesses ready to expand have faced multi-year waits and uncertain costs. For a large operator with deep pockets and a long horizon, that is a manageable frustration. For a small firm trying to add a production line or electrify a site, a long, expensive, uncertain connection can simply kill the project.
If hyperscale sites are allowed to absorb the available headroom, the connection queue for smaller firms gets longer, not shorter. The framework's argument is that grid reinforcement in and around Growth Zones must be planned strategically, so that it does not prioritise the largest sites at the expense of local demand. Growth Zones should be a reason to strengthen the grid for everyone in the area, not only for the anchor tenants.
Reserving room for local demand
One practical step is to reserve a clear share of new or reinforced local capacity for SMEs and public services, rather than allowing it all to be committed to a small number of very large users. Knowing that a portion of capacity is protected gives smaller firms the confidence to plan investment.
Flexible connection offers matter too. Not every firm needs firm, always-on capacity at maximum draw. Flexible or time-of-use connection arrangements can let more firms connect sooner and at lower cost, in exchange for managing their demand at peak times. For many SMEs that is a perfectly acceptable trade, and it makes better use of the network. These options should be actively offered and explained, not buried in technical complexity that only large firms have the resources to navigate.
Anchoring renewables for shared benefit
The demand created by AI infrastructure can also be used to anchor new renewable generation and storage. A large, predictable load is exactly the kind of customer that makes new solar, wind and battery projects viable. If that generation and storage is planned so it also serves surrounding firms and communities, the Growth Zone becomes a catalyst for local clean power rather than simply a consumer of it.
That is the difference between a data centre that draws power out of a region and one that helps build lasting energy capacity into it. Community benefit from Growth Zone development can and should include energy: reinforcement that unlocks local connections, renewables that lower long-term costs, and storage that improves resilience for more than just the anchor site.
The underlying point is simple. If Wales wants AI Growth Zones to raise productivity across the real economy, then the firms in that economy need to be able to plug in. Grid access is not a technical footnote to the AI story. For smaller Welsh firms, it is often the difference between being able to adopt AI and being shut out of it.