The AI Race's Forgotten Bottleneck: UK Executives Warn 5G Upgrades Are Too Slow to Carry AI Traffic

The global AI infrastructure race has, until now, been framed almost entirely as a datacentre story: gigawatt campuses, grid connections, water for cooling, and the GPU fleets inside. But in a Guardian interview published this morning, senior UK telecoms executives argued that the race has a second, largely ignored front — the mobile and broadband networks that will actually have to carry AI traffic to hundreds of millions of devices. Their warning is blunt: if the UK does not accelerate its 5G upgrades, it risks becoming a laggard in the AI era regardless of how many datacentres it builds.

The uncomfortable baseline

The executives’ case rests on an awkward foundation of evidence. According to a July analysis by consumer group Which?, based on data from network analytics firm Opensignal, mobile coverage in the UK is worse than in any of the 27 EU member states and every other G7 nation. The UK ranks 57th globally for overall network performance, 70th for download speeds, and 55th for the consistent quality needed for activities like video calls, streaming and gaming.

Britain has already fixed the fixed-line half of this problem. After years of embarrassment over having some of Europe’s slowest broadband, BT’s Openreach and other providers have blanketed the country with full-fibre connections. The mobile side is where progress has stalled — precisely the half of the network that matters most for the next wave of AI devices.

“If with the technology and services we use today [networks] are sub-optimal, imagine when you add mass AI usage and other associated activities,” said Andrea Donà, networks director at VodafoneThree, the UK’s largest mobile operator, formed from the £16.5 billion Vodafone–Three merger. “I would go further. I am embarrassed today at the level of quality of our networks. The UK is worthy of what EE, O2 and VodafoneThree can offer, but we need reform.”

Why AI changes the calculus

The core of the executives’ argument is a capacity curve. AI adoption is shifting from occasional chatbot queries to continuous, ambient usage — voice agents, real-time video understanding, on-device models syncing with cloud inference, and agentic workflows running in the background of everyday applications. Every one of those patterns multiplies upstream and downstream traffic on mobile networks.

Donà offered a vivid proxy for what constant AI usage will feel like on an under-provisioned network: the signal collapse people experience at mass gatherings. “What we see now only with a mass aggregation of people — a gig at Wembley, the British Grand Prix at Silverstone, a rugby international at Twickenham — is a proxy, an indication of the things that will come,” he said. “That one-off experience will be constant because of the additional capacity consumers will need in an AI world.”

Another senior telecoms executive, speaking to the Guardian, put the timeline at two to three years: “Things are going to change massively with the next wave of devices, in two or three years when AI is integrated into everything. AI companies, and governments, don’t think about networks — but they will when they hold them back.” He predicted that AI connectivity will become a “national competitiveness issue” in the same way the slow full-fibre broadband rollout did a decade ago.

The £11 billion investment stuck in planning

The operators are not standing still. VodafoneThree has committed £11 billion to build what it calls the UK’s best 5G network, targeting 5G standalone (SA) coverage of 99% of the population by 2030 and 99.96% by 2034. The company has already upgraded 8,000 sites as part of the programme and completed what it describes as a world-first integration of the two merged networks onto a single core.

The bottleneck, executives insist, is not capital or ambition — it is planning law. Donà points out that VodafoneThree is actually reducing its nationwide mast count by 30% as it consolidates the merged estate; the challenge is the restrictive rules around upgrading the sites that already exist. About 96% of the upgrades VodafoneThree needs are on existing sites, yet more than half of those will require some form of planning permission — a process the industry describes as lengthy, inconsistent and bureaucratic. Industry figures cited earlier this year put planning rejection rates as high as 86% in some areas, and an analysis published by Politico in May estimated that cutting planning red tape around 5G SA upgrades could be worth a £2.58 billion boost to the UK economy.

“Planning reform is the lowest possible cost growth policy — it doesn’t require any government money,” Donà said, noting that infrastructure investment is ready to support elected leaders’ growth agendas and calling it “a great first 100 days win for the new prime minister.”

There is also a fairness argument. Mobile networks are legally classified as critical national infrastructure — the same category as datacentres, energy and water systems — yet operators say they receive none of the business-rates and energy-cost reliefs extended to those sectors. The operators’ implicit message to government: if connectivity is strategic infrastructure for the AI era, treat it like one.

The bigger picture

The warning lands at a moment when AI policy debate in the UK has been dominated by compute, chips and safety regulation. The telecoms argument reframes the problem end-to-end: an AI economy is only as strong as the full pipeline from datacentre to device, and the last wireless mile is becoming the constraint. Other countries are moving faster — the operators note that the UK’s 57th-place ranking places it behind nations with far smaller economies, and every G7 peer.

There are caveats. Network operators have obvious commercial incentives to push for planning deregulation and tax relief, and the Wembley-style crunch analogy assumes AI usage patterns that are still emerging. Whether ambient AI devices will stress networks as heavily as forecast depends on how much inference moves on-device versus to the cloud — a shift that would cut traffic in both directions.

But the direction of travel is not seriously disputed. AI traffic of some substantial magnitude is coming, 5G standalone is the infrastructure designed to absorb it, and the UK’s rollout is measurably behind its peers. The executives’ core point is about sequencing: networks take a decade to build, and the devices arrive in two to three years. Waiting until the traffic arrives to fix planning rules means spending the AI decade’s first years throttled — a familiar story for anyone who watched the UK’s late dash to full-fibre broadband.

For a country positioning itself as an AI heavyweight, the message from the people who run its networks is simple: the AI race will be decided not just by who builds the biggest datacentres, but by who can actually connect everyone to them.