Wales should welcome major AI infrastructure. Data centres matter. They can attract investment, create construction work, improve national visibility and help place Wales inside the next wave of AI development. But treating data-centre announcements as the same thing as economic transformation would be a serious mistake.
The real question is not whether Wales can host more AI infrastructure. It is whether that infrastructure changes the performance of the wider Welsh economy.
That distinction matters because Wales is now at a genuine point of leverage. Two AI Growth Zones, one in North Wales and one in South Wales, are expected to attract major private investment and large energy demand. At the same time, the Welsh Government’s AI plan puts data centres, adoption hubs and AI skills at the heart of a wider economic story. That creates momentum. It also creates risk: once the headlines land, it becomes easy to confuse enabling infrastructure with delivered economic value.
Infrastructure is an enabler, not the outcome
Data centres are important, but they do not automatically spread prosperity on their own. Many of the biggest employment effects come during construction and site build-out. The longer-term economic test is different. It is whether the surrounding economy becomes more productive, more capable and more competitive because those assets are there.
For Wales, that test is unavoidable. The country still faces a long-running productivity challenge, and small and medium-sized enterprises account for around 62.3% of Welsh employment. If most people work in SMEs, then the route to better economic performance cannot run through a handful of large sites alone. It has to run through the firms that make up everyday Welsh business life.
A data centre can sit in Wales without materially changing how a food producer in Carmarthenshire operates, how a manufacturer in Newport plans production, or how a care provider in Ceredigion manages staff time and reporting. That is why infrastructure alone is not enough.
The enclave risk
The danger is not that Wales gets data centres. The danger is that Wales gets only data centres.
If Growth Zones become fenced-in enclaves, then the benefits may concentrate around land, construction, grid access and a relatively small number of specialist roles, while the wider economy remains stuck with the same barriers it had before: weak AI adoption, limited access to trusted providers, patchy practical support, and long waits for energy connections.
That outcome would still produce impressive press releases. It might still produce large investment figures. But it would not answer the core economic question. Wales does not need infrastructure that looks transformative while leaving the SME base largely untouched. It needs infrastructure that becomes a platform for wider diffusion.
That means asking harder questions. Are major projects buying from Welsh suppliers? Are smaller Welsh firms gaining access to contracts, skills and specialist work? Are businesses outside the site boundary getting practical help to adopt AI in ways that improve output, quality or resilience? Are local energy and grid plans being shaped so that smaller firms are not pushed to the back of the queue?
If the answer is no, then Wales may be hosting strategic infrastructure without capturing enough of its long-term value.
What success should look like
A better model starts by treating data-centre investment as a means, not the end in itself.
First, public support for major AI infrastructure should come with visible local conditions. That means supplier targets, transparent reporting on Welsh SME spend, and skills programmes that help smaller firms and local workers participate, not only the large operators.
Second, AI adoption funding should be built around real work inside real Welsh firms. Not one-off pilots. Not generic awareness sessions. The firms that matter most need several months of practical support to redesign workflows, test use cases, train staff and measure results.
Third, grid and energy planning has to work for the wider economy. If Growth Zones absorb capacity so completely that SMEs, public services and small industrial users face long delays or higher costs, then the infrastructure will end up deepening the very barriers Wales needs to remove.
The deeper point is simple. Wales will not solve its productivity challenge by hosting compute alone. It will do so by turning infrastructure into adoption, adoption into capability, and capability into wider economic gains.
More than racks and substations
The strongest case for AI Growth Zones is not that Wales can host more racks. It is that Wales can use this moment to build stronger local supply chains, more capable SMEs, better routes into skilled work and a broader base of specialist Welsh firms. That is a harder standard, but it is also the only one that makes the long-term economic case credible.
Data centres matter. Wales should want them. But the real prize is not simply being a place where AI infrastructure is built.
It is becoming a place where that infrastructure helps the wider Welsh economy perform better.