Site icon vXtream

Hyperscale Data Centres: Assembling J.A.R.V.I.S. – But at What Cost?

Image-of-J.A.R.V.I.S-type-interface-to-illustrate-vXtream-article-on-AI-systems-and-the-issue-of-hyperscale-data-centres

The UK needs more data centre capacity to support the next generation of digital services. But as the Government rolls out the welcome mat for a new generation of AI infrastructure – hyperscale data centres – are we asking enough questions about what we’re building, why we need it and who ultimately pays the price?

If you’ve watched the Marvel Studios’ Iron Man or Avengers films, you’ll probably be familiar with J.A.R.V.I.S. (Just A Rather Very Intelligent System). Tony Stark’s ever-present digital assistant manages his home, controls the Iron Man suits, monitors security and generally makes life considerably easier for a billionaire inventor.

Named after the Stark family’s butler, Edwin Jarvis, it’s an impressive bit of technology. And, if we’re honest, quite a useful personal assistant to have around.

But while Hollywood’s interpretation of artificial intelligence (AI) might involve sophisticated computer systems, autonomous robots and futuristic laboratories, the reality of AI in everyday life is rather less spectacular.

It’s asking ChatGPT to help write a letter. It’s Siri answering a question, Alexa playing music or Google Maps finding a quicker route around a traffic jam. It’s Netflix suggesting something to watch, an online retailer recommending a product or a bank identifying a potentially fraudulent transaction.

AI is already woven into many of the services we use every day. We might not always recognise it, understand how it works or even realise we’re using it. And that’s perhaps part of the problem.

We’re hearing an awful lot about AI. Governments are talking about becoming AI superpowers, technology companies are investing billions in computing infrastructure and developers are proposing enormous new hyperscale data centres to support the next generation of intelligent systems.

But how much of this do we really understand? More importantly, does the public understand why these developments are happening and what they mean for the communities expected to accommodate them?

The infrastructure behind the intelligence

AI doesn’t simply exist in the cloud. Behind every online service, digital assistant and increasingly sophisticated AI application is a physical infrastructure of servers, storage, networking and power.

Data centres provide the environment in which this infrastructure operates. They store information, process requests and deliver the computing resources that underpin everything from online banking and public services to business applications and AI.

As demand for digital services increases, so does the requirement for the infrastructure supporting them. And the numbers suggest that the UK is already facing a capacity challenge.

According to Savills’ Spotlight: UK Data Centre Building Capacity – H1 2026, available UK data centre capacity is becoming increasingly constrained. Vacancy has fallen to just 8%, dropping to 7% in London. Meanwhile, the UK recorded a record 217MW of data centre take-up during 2025.

The UK now has approximately 1,803MW of live data centre capacity, but London accounts for around 91% of that total. Its established connectivity, cloud infrastructure and network density make it an attractive location, but also place considerable pressure on available power, land and development opportunities.

The challenge isn’t simply building more facilities. It’s finding locations where the necessary electricity, connectivity and supporting infrastructure can actually be delivered.

The Government estimates that the UK will need 6GW of AI data centre capacity by 2030 if it is to maintain its position in the international AI market. Meanwhile, Ofgem has reported that developers are seeking electricity connections totalling around 72GW. For perspective, the UK’s peak electricity demand last year was approximately 45GW.

These figures aren’t directly comparable: the 72GW represents requests in the connection pipeline, not confirmed demand or a prediction that every proposed project will proceed. Nevertheless, it illustrates the extraordinary scale of competition for access to the national grid.

The question of whether we need additional data centre capacity is therefore becoming increasingly difficult to avoid. But that doesn’t automatically mean we need every proposed development.

Not all AI is created equal

One of the difficulties with the current debate is that AI is frequently treated as a single technology with a single set of infrastructure requirements.

It isn’t.

Consider the difference between asking ChatGPT to summarise a document and training an entirely new AI model using enormous quantities of information and computing power. Both involve AI, but the resources required are very different.

AI training involves feeding vast amounts of data into a model, allowing it to identify patterns and develop the ability to perform particular tasks. This can require thousands of specialised processors operating together, consuming enormous amounts of electricity.

AI inference is what happens when that trained model is actually used. It’s the process behind generating an answer, recognising an image, making a prediction or producing a recommendation.

As AI becomes increasingly embedded in everyday business and consumer applications, demand for inference computing is likely to grow. Unlike some of the enormous training operations, these workloads may be distributed across a wider network of facilities, including smaller deployments closer to businesses and users.

Savills highlights this distinction in its research, suggesting that while large-scale training campuses will have a role, inference could become an increasingly important driver of future infrastructure requirements.

It’s an important distinction. We shouldn’t assume that every application of AI requires the functionality of hyperscale data centres, any more than every business needs its own power station. Understanding what we’re building and what it will actually be used for should be part of the conversation.

The price of progress

There is, however, another side to the story. The race to build increasingly powerful hyperscale data centres is creating some very real challenges for the UK’s energy infrastructure and environmental ambitions.

The BBC’s File on 4 Investigates recently examined the potential consequences of the Government’s ambition to accelerate AI data centre development.

Its investigation identified four proposed developments considering gas-fired power generation while waiting for connections to the national grid. BBC calculations suggested that, if operating continuously at full capacity, these facilities could generate approximately 4.1 million tonnes of carbon emissions annually. That’s more than the combined emissions of Manchester, Liverpool and Southampton, according to the investigation.

That’s a particularly uncomfortable prospect for a country committed to achieving net zero emissions by 2050.

The Government has said that exceptionally strategic data centre projects could be permitted to generate electricity on-site while waiting for grid connections, subject to tight controls. The intention is to prevent delays to strategically important developments.

But there is an obvious question: if temporary fossil-fuel generation becomes a convenient long-term solution, are we simply creating another environmental problem in the pursuit of technological progress?

Electricity isn’t the only concern.

Cooling large computing facilities requires careful management of heat and, depending on the technology employed, significant quantities of water. The cumulative impact on local water supplies, air quality, noise and the surrounding environment cannot simply be dismissed because a development promises investment and employment.

And these aren’t abstract concerns. In Scotland, proposed hyperscale data centres have already prompted questions about renewable energy commitments, reliance on the national electricity grid and the cumulative impact of additional demand.

The Scottish Government has responded by requiring Environmental Impact Assessments for all new data centre developments exceeding 50MW of power capacity. It has also introduced requirements for planning authorities to notify Ministers of relevant applications.

The debate is clearly moving beyond the technology industry.

The growing backlash: not in my backyard?

The argument about the expansion of hyperscale data centres is no longer confined to planning committees, energy regulators and technology companies. Increasingly, it is taking place in the communities expected to accommodate these developments.

In Amsterdam, activists recently broke into a Microsoft data centre construction site, climbing cranes and confronting police. The incident formed part of a wider European campaign against large-scale data centre developments, with environmental groups and digital rights campaigners questioning the speed and consequences of the industry’s expansion.

The protests reflected a mixture of concerns, including electricity and water consumption, environmental damage, the influence of major technology companies and the wider implications of increasingly powerful AI systems.

It would be wrong to suggest that the actions of a relatively small number of activists represent public opinion. But they do illustrate an emerging problem: data centre expansion is becoming a wider social and political issue, rather than simply a planning and infrastructure matter.

There are also practical questions about whether the infrastructure required to support these developments is actually available.

In the UK, Microsoft’s withdrawal from its involvement in Nscale’s proposed £2bn AI data centre in Loughton, Essex, provides a particularly interesting example.

The project was intended to support a major AI supercomputer, with Microsoft providing the funding for its computing infrastructure. However, the expected electricity connection has been pushed back from the end of 2026 to the 2030s. Microsoft subsequently terminated its involvement as the project’s anchor customer, although Nscale maintains that it intends to proceed and is exploring alternative power sources.

The proposed development is expected to create around 250 permanent jobs when operational. It’s a reminder that investment announcements and development proposals are not the same as operational infrastructure. Without sufficient electricity, connectivity and supporting services, even the most ambitious project can struggle to become reality.

Environmental concerns are also attracting international attention.

In Finland, the Guardian reported that approximately 250,000 trees had been cleared from a 330-hectare site near Muhos in preparation for a Google data centre development. The Finnish authorities ordered preparatory work to stop following concerns about the way the land had been cleared.

Local residents and environmental campaigners questioned the impact on water quality, wildlife and the surrounding landscape. Google said it had stopped preparatory work following the authorities’ intervention and outlined plans for environmental mitigation and tree planting.

These examples are not necessarily representative of every development or every community. Hyperscale data centres can bring investment, employment and valuable infrastructure. But they demonstrate that the industry cannot simply assume that communities will welcome development because it carries an AI label or promises economic growth.

There is a growing need to establish what local communities actually receive in return, how environmental commitments are monitored and whether the benefits genuinely justify the demands placed on local infrastructure.

The industry needs more than planning permission and a grid connection. It needs public confidence. And that is something that cannot be secured simply by telling people that AI is the future.

A Covenant with the public

Against this background, the UK Government’s announcement of its Data Centre Covenant at the end of October’s Autumn Budget is significant.

The intention is straightforward: accelerate the development of strategically important hyperscale data centres while ensuring that developers take greater responsibility for the consequences of their investment.

The Covenant sets out five broad expectations.

Developers should pay their fair share of the energy costs they create, invest in the communities hosting their facilities, meet high environmental standards, create good-quality jobs and apprenticeships, and strengthen Britain’s digital security and resilience.

There are some welcome commitments.

New facilities are expected to achieve a Power Usage Effectiveness (PUE) of 1.3 or better, minimise their reliance on public water supplies and consider capturing and reusing waste heat where practical.

Strategic developments will also be expected to publish annual information on energy and water consumption, emissions and efficiency. Local Benefit Plans are another important element, intended to ensure that communities hosting nationally significant developments receive meaningful investment in return.

However, there is an important distinction between an expectation and an enforceable obligation.

The Covenant encourages all data centres to meet its tests, but strategic developments must satisfy them to qualify for additional Government support. The precise definition of a strategic project, how commitments will be assessed and how compliance will be monitored and enforced still require further clarification.

BCLP’s analysis also highlights the possibility of a two-tier market emerging, in which strategically designated developments receive preferential access to electricity connections and other Government assistance, while smaller or non-strategic projects do not enjoy the same support.

The Government has signalled its intentions. Turning those intentions into measurable outcomes is the next challenge. A covenant is only as good as the commitments it creates, the evidence it demands and the consequences when those commitments aren’t met.

Bigger isn’t necessarily better

The UK has a genuine opportunity to become a leading location for digital infrastructure and AI development. But success shouldn’t be measured exclusively in gigawatts, acres of land or the size of the next hyperscale campus.

There is a wider infrastructure conversation to be had. Not every organisation needs hyperscale computing resources. Not every AI application requires thousands of GPUs. And not every business needs to put all its infrastructure into a single cloud environment.

Efficient, interconnected regional data centres, appropriately sized computing resources and better utilisation of existing infrastructure all have a role to play.

The ability to locate workloads in the right environment, close to users where necessary, while maintaining resilience, security and control, is every bit as important as the headline capacity figures. For businesses, that means having choices. For infrastructure providers, it means delivering the right combination of performance, connectivity, security and efficiency rather than simply building bigger.

It also means recognising that digital infrastructure is not an isolated technology investment. It is part of the UK’s wider energy, economic and environmental landscape.

From artificial intelligence to intelligent infrastructure

Let’s return to J.A.R.V.I.S. for a moment.

Tony Stark’s assistant is impressive because it makes complex technology useful. It quietly manages an extraordinary amount of infrastructure, allowing Stark to concentrate on what he wants to achieve. That’s perhaps a reasonable ambition for the real world, too.

AI should make our lives easier, help businesses become more productive, support scientific and medical advances and improve the services on which we depend. But the infrastructure supporting that ambition must be developed intelligently.

We need sufficient data centre capacity to support our digital economy, protect critical services and give British businesses access to the technologies shaping their future. We also need to ensure that investment delivers genuine benefits, rather than transferring the costs and consequences to the communities and energy consumers expected to accommodate it.

The Government’s Data Centre Covenant is a welcome recognition that these responsibilities matter. But words on paper won’t build a sustainable digital economy. That will require transparency, accountability, sensible planning and a willingness to question whether every proposed development represents the right investment.

We don’t need to build the next generation of data centres simply because we can. We need to build them because they serve a purpose, because they deliver value and because we can do so responsibly.

Perhaps that’s what intelligent infrastructure should really mean.

Want to know more?

If you’d like to discuss the issues raised in this article, or find out more about how vXtream delivers secure, sustainable and responsible infrastructure, let’s talk about your requirements.

Contact Us Now.

J.A.R.V.I.S. (Just A Rather Very Intelligent System) is a fictional artificial intelligence assistant from Marvel’s Iron Man universe. Marvel and related character names are the property of their respective rights holders. This article references the character for illustrative purposes and is not affiliated with or endorsed by Marvel or Disney.
Image supplied by Magnific

Exit mobile version