Planning reform may be helping accelerate data centre development, but Lee Myall, CEO of Neos Networks, argues that connectivity risks becoming the next major barrier to the UK’s AI ambitions.

The UK’s AI data centre boom might have been amassing headlines for a while now, but it’s clear to me that the government is only recently starting to get serious about creating the conditions needed to support large-scale digital build-out.

You only have to look at the reforms to the Nationally Significant Infrastructure Projects (NSIP) regime that came into force on 24 July 2026. Ministers have officially scrapped mandatory pre-application consultation for NSIPs, stripping months off planning timelines for data centres and potentially cutting costs by £1 billion across this Parliament. So far, three data centre proposals – in Buckinghamshire, Bedford and Dartford – have already been directed into the NSIP regime, and early signals from Andy Burnham’s agenda suggest these reforms will stay in place.

But at the same time, while planning is speeding up, national policy is becoming slightly murkier, with the AI Growth Zones programme being a case in point. After the closure of the Department for Science, Innovation and Technology (DSIT) in a wider Whitehall reshuffle, the scheme has effectively been left in limbo. Officials say they remain committed to delivering the zones, but there is still no clarity on whether the programme will be reshaped or whether the team originally tasked with delivering it will stay in place.

The result is a volatile environment for developers and investors: they are being encouraged to build faster, but on uncertain political ground. Accelerating planning approvals while key policy details remain up in the air only increases the danger of building in a vacuum. A site can be approved, funded and announced long before it is genuinely ready to operate, and when that happens, the industry tends to discover, often too late, that the missing piece is not land or even power, but connectivity.

Lessons from the 2000s fibre boom

I keep thinking about the early 2000s, when operators and investors poured money into fibre on the assumption that demand would appear. It didn’t – at least not straight away – and so what followed was a painful period of stranded assets, empty capacity and a telecoms market that spent years working through the consequences.

Don’t get me wrong, I don’t think we are about to repeat that cycle in exactly the same way. It’s unlikely that we’ll build ahead of demand again because the signal from AI is real and far more immediate than the speculative optimism that shaped that earlier boom. What worries me now is sites getting approved, funded and announced, but not being built on time because of fibre availability.

In fact, recent research shows that 82% of UK data centre operators said they had already delayed site builds or expansions because of fibre availability, while 45% of enterprises see fibre as the key bottleneck holding back AI and digital infrastructure.

Compute follows power and land, but it still needs fibre

For years, the default map of UK data centre growth was concentrated in and around London, especially places like Docklands and Slough. But the government’s AI Growth Zones programme is broadening the regions for data centre development, with locations such as Oxfordshire, the North East, North Wales, South Wales and Lanarkshire.

This decentralisation is a positive thing because it can help align compute more closely with power, land and regional opportunity. But at the same time, it creates a challenge in regions where there’s a lack of underlying fibre routes, ducts and wayleaves because these factors were not prioritised in the first place when choosing sites. To help avoid these delays, fibre planning needs to take place alongside data centre development, particularly in locations where AI-related demand is visibly emerging. That includes considering how new clusters connect with established data centre hubs, cloud interconnection points and internet exchanges, while avoiding the kind of speculative overbuild seen in the past.

What’s been interesting over the past few months is the change we’re starting to see in how fibre is being approached. Data centre developers and operators are slowly getting involved much earlier in discussions around fibre availability and delivery timescales, while also asking for much more capacity from day one and planning for a far steeper ramp-up in demand.

That suggests two things to me. First, expectations around take-up and data traffic have clearly moved on. And second, there’s a bigger push to de-risk developments by gaining confidence in fibre and connectivity much earlier in the process.

AI infrastructure players want fibre that can keep pace with deployment

So, say you’ve got a site with planning consent in place and a reliable fibre backbone already available. Job done? Not quite. Network operators also need to reckon with another factor: how quickly and predictably that connectivity can be delivered.

Hyperscalers and neocloud providers still purchase dedicated fibre services, but their deployment timelines are accelerating. At the same time, when purchasing internet services like Direct Internet Access (DIA), they expect to order and scale capacity dynamically, much like software. Across both models, network providers are under growing pressure to overcome operational constraints such as wayleaves and speed up deployment, giving customers greater confidence that connectivity will be delivered on schedule.

AI workloads are also starting to shape future rack-density designs, with some moving towards 1 MW. That brings huge pressure not only on power and cooling, but also on fibre density and how quickly capacity can be delivered.

As these AI infrastructure companies continue to scale, I think network providers will increasingly need to offer fast and predictable provisioning, clear visibility of available routes and diverse network paths that minimise disruption if a route is affected. These qualities are becoming increasingly important as the UK pursues its sovereign AI ambitions, ensuring the underlying infrastructure is secure, reliable and able to support critical workloads over the long term.

Having fibre in the right place may be enough to get you into the conversation, but increasingly it will be how easy that fibre is to access and use that determines whether a development can progress as planned. And that’s the final puzzle piece worth paying attention to as the UK races to build out AI infrastructure. Planning reform and power availability might get a site off the ground, but it’s connectivity, and how easily it can be consumed, that decides whether it actually delivers.