Nokia and Virgin Media O2 Achieve 20% Spectral Efficiency Gain With AI-RAN on Live 5G

Nokia's AI-RAN technology delivered a 20% improvement in spectral efficiency on Virgin Media O2's live commercial 5G network — not in a controlled lab environment, but across real subscriber traffic on a production network serving millions of U.K. customers.

The result matters because it validates AI-powered radio access network (RAN) optimization at the scale and conditions that count. Lab benchmarks for AI-RAN have been circulating for two years. A validated gain on a live carrier network serving real traffic is a different and more meaningful proof point for the rest of the industry.

What the result means

Spectral efficiency measures how much data a network can transmit per unit of radio spectrum — essentially, how well the network uses the airwaves it has. A 20% improvement in spectral efficiency means O2 can serve 20% more data capacity from its existing licensed spectrum, or maintain the same capacity while using 20% less spectrum — both of which translate to either better service quality for subscribers or reduced operating costs for the carrier.

According to Telecoms Tech News, the efficiency gain was validated across O2's live network, with Nokia's anyRAN software stack handling the optimization layer. The AI system continuously analyzes network conditions — signal quality, interference patterns, subscriber density, traffic load — and adjusts how spectrum is allocated across radio resources in real time.

Why traditional RAN can't do this

Conventional radio access networks use static configuration rules and scheduled optimization cycles. Engineers set parameters based on predicted traffic patterns and update them periodically. The network doesn't adapt dynamically to changing conditions between optimization cycles.

AI-RAN replaces static rules with continuous inference. When a large crowd suddenly appears in one cell sector and disperses from another, the AI system reallocates radio resources in real time rather than waiting for the next scheduled configuration update. When atmospheric conditions change propagation characteristics, the system adjusts without a field engineer deploying a configuration change.

The result is a network that uses its spectrum more efficiently under variable conditions — which is most of the time, on a network serving millions of subscribers with unpredictable behavior patterns.

Industry adoption accelerating

Nokia's anyRAN software stack now has approximately 20 operator deployments globally, with more announced quarterly. The deployment with O2 joins a growing list of European carriers validating AI-RAN in live environments, following results from operators in Germany, Sweden, and Finland that reported similar efficiency gains in earlier trials.

The competitive dynamics in telecom are pushing adoption faster than operators originally planned. A carrier that deploys AI-RAN and gains 20% more usable capacity from existing spectrum holds a structural advantage over competitors running static network configurations — especially in dense urban markets where licensed spectrum is the most constrained resource.

Nokia is not the only vendor in the AI-RAN space. Ericsson, Samsung Networks, and several software-first vendors including Viavi and Parallel Wireless are shipping competing products. But Nokia's combination of anyRAN's software stack and NVIDIA's GPU-accelerated compute infrastructure — used for the AI inference that drives the optimization — gives it a full-stack offer that carriers can deploy without integrating components from multiple vendors.

What to watch

The efficiency gain demonstrated at O2 establishes a benchmark that other carriers and vendors will now be measured against. The near-term question is whether the 20% gain holds as network load increases — high-efficiency gains are easier to achieve at moderate load conditions, and harder to sustain at peak congestion when the optimization problem becomes more constrained.

Nokia and O2 haven't published the specific traffic load conditions under which the 20% figure was measured. Carriers considering AI-RAN deployments should ask for disaggregated performance data across load bands before committing to large-scale deployments.