VMO2 and Nokia claim 20% spectral efficiency gain from AI in live 5G RAN trial

Virgin Media O2 recorded a roughly 20% improvement in spectral efficiency by applying AI to its 5G radio access network during a live trial with Nokia at London Paddington rail station, an experiment the UK operator describes as the first of its kind in the country.

The trial site serves customers around one of the capital's busiest transport hubs. Software predicted and optimised 5G radio resources as network conditions and demand levels shifted through the day. The system then allocated capacity where the traffic actually was, rather than where a static plan expected it to be.

What did the AI actually change?

The technology let VMO2 carry more data over the spectrum it already holds, the operator said, and delivered what it called "significant throughput gains" during periods of high demand. In practice, that means faster downloads and higher-quality streaming in crowded locations, with less buffering when the station fills up.

VMO2 CTO Jeanie York framed the result as a response to relentless traffic growth rather than a showcase for its own sake. "As demand for data keeps growing, we need smarter ways to make the most of the spectrum and infrastructure we already have," she said.

"This live trial with Nokia showed how AI can improve network efficiency and help us respond more effectively to changing customer demand," York added. "By making better use of existing network resources, we can deliver a better experience for customers while supporting future traffic growth."

That last point carries commercial weight. Spectrum in the UK is expensive and finite, and operators face continued traffic growth from video, gaming and fixed-mobile convergence. A 20% efficiency gain on existing holdings is, in effect, capacity the operator did not have to buy at auction.

Vendor claims versus live-network evidence

Nokia president of mobile networks Emma Falck said the test "demonstrates how AI is moving from experimentation to delivering measurable benefits in live mobile networks". The collaboration, she added, "shows the significant potential of AI-native networks to enhance customer experiences while improving operational efficiency".

Falck's second statement is a marketing claim about future potential; the first rests on measured results from a single live site. The distinction matters in a segment where AI-RAN announcements have multiplied faster than verified deployments.

Where the trial sits in a wider plan

VMO2 positioned the Paddington experiment as part of an ongoing network transformation plan with Nokia, rather than a one-off demonstration. The operator has been working through a broader upgrade of its mobile infrastructure, and AI-driven radio optimisation fits that trajectory.

The timing also connects to a wider industry push. The trial comes a week after a series of partner announcements involving Nokia's work in AI-RAN alongside Nvidia, as vendors argue the technology's momentum is justified by results rather than hype.

For now, the Paddington figures give that argument its most concrete UK data point. Whether VMO2 moves from a single-site trial to wider deployment across its network, and on what timeline, will show whether the 20% figure scales beyond one busy station.