Prefabricated data centres: A scalable solution to take on the AI age

AI represents the most consequential technological shift in human history. For the first time, we are manufacturing intelligence itself, which is driving the largest mobilisation of capital the world has ever seen. Forecasts are suggesting that global data centre capacity could rise by as much as 200GW by 2030 – a significant increase on today’s footprint.

There is no denying that the UK has stepped up to the challenge of the century with a plan to unlock heavy investment in building out AI infrastructure. The reality is we cannot scale as quickly as we need using traditional methods and we must think in different and innovative ways. Fortunately, there is great reason for optimism as new platforms are due to enter the market shortly that will build at a scale that no one has previously envisioned.

We’re already seeing the proliferation of value-adding AI use cases across every sector, and this is reshaping what a data centre needs to deliver in the future. AI data centres will be AI factories, as these new workloads demand radically different infrastructure from the facilities of the past. At the same time, escalating energy demands, supply chain bottlenecks and global skills shortages are putting traditional build models under pressure.

The good news is that building better infrastructure for AI is a big priority. However, it’s important we understand how the industry can design, implement and scale the data centres of the future, tackling AI-based challenges and other issues. Part of the answer is undoubtedly an industrialised and prefabricated approach to building, particularly for many large-scale deployments around the world.

Two challenges for data centre deployment

The data centre industry is currently up against two key challenges which are reshaping how capital is deployed and the timeline to deliver returns —shortage and speed. We’re seeing a significant shortage of talent across both construction trades and critical infrastructure roles, worsened by the sheer demand of AI. This talent gap is inflating project costs and delaying project delivery, even in cases where capital is readily available. At the same time, supply chain volatility continues to drive price increases and equipment shortages, shifting the bottleneck upstream.

Importantly, there’s also a shortage of power to meet the explosive demand for AI compute. The demand for electricity is outstripping the available supply. As an industry, we will need to look at new and different types of power sources to bridge the gap. We need to unlock innovative ways to power these sites to get GPUs up and running as quickly as possible.

Speed and schedule certainty is a well-known issue in the industry, and AI has held up a microscope to the fact that progress simply isn’t going fast enough. New plans for data centres in the UK are making daily headlines right now, although these are not anywhere near nor quick enough for what is needed. Yet, building large-scale new sites in a traditional manner will typically take two to three years, and sometimes even longer if there are labour and equipment shortages. However, the AI industry does not have time right now. We need a new way.

The potential of prefabrication

To address these key challenges, the industry must look to industrialised delivery mechanisms. Standardisation and offsite manufacturing are no longer niche; they are essential tools to de-risk capital investment and compress deployment cycles.

Prefabrication enables scalable and repeatable quality, reduces the need for on-site resources, and accelerates the time to first token and revenue generation. For institutional investors and infrastructure funds, this approach introduces greater clarity and speed into capital planning, offering a path to scaled deployment without the typical trade-offs in quality or timeline.

In addition, prefabricated solutions deliver greater schedule certainty, which is a key factor that investors and operators must consider. Manufacturing offsite in controlled environments shortens timelines, reduces risk and bypasses many local constraints, including skilled labour shortages. It allows data centres to be treated as ‘products’ rather than long-term conventional construction projects, reducing timelines from years to just months from conception to full deployment.

Why can’t legacy sites just adapt?

It’s not just speed that can be tackled with prefabrication, but also power shortages. Legacy sites were never designed to handle the current power demands, and this issue is only going to increase if nothing changes.

While liquid cooling is becoming standard, this requires intricate planning and infrastructure that older sites rarely accommodate well. Retrofitting is possible but often costly, disruptive and inefficiently created. Prefabricated data centres that are purpose built for GPUs can integrate advanced cooling, optimise power and support high load densities without compromises.

There’s also the opportunity to ‘follow the power’ if a data centre is manufactured offsite e.g., easily deploying data centres in an area where there is more grid capacity. Location flexibility enables energy-efficient deployments that can be positioned closer to available renewable energy sources, reducing both grid strain and environmental impact.

Looking ahead

It’s clear the industry is at a crucial inflexion point. Amid the exponential growth of AI and soaring workloads, inference costs are falling, use cases are multiplying and demand for capacity is set to accelerate. What is built today needs to be able to overcome the challenges of tomorrow. Therefore, facilities should be designed for easy upgrades to handle shifting demands, like new cooling systems, AI workloads and shorter GPU lifecycles. And it is the responsibility of both the developers and the stakeholders to prioritise education and understanding about what these future needs will look like for their digital infrastructure.

GPU-infrastructure lies at the heart of the transformation taking place across every industry, and prefabricated data centres serve as a solution to many of the strains being put on infrastructure today. If the UK is to live up to its potential as a ‘superpower’ for AI technology, infrastructure cannot be compromised. Introducing scalable, prefabricated solutions delivers an essential pathway for futureproofing data centres.