StocksMedium•20 November 2026•
4 min read

Vodafone Three Plans AI Network Tests, Targets Speeds of 15 Mbps

By

Key Facts

1Vodafone Three announced the partnership on November 18, 2026, for trials planned in the UK.
2The SuperMobile example offered speeds of up to four times those of standard plans in congested areas and a guaranteed minimum of 15 megabits per second.
3Spectrum efficiency improved by about 20% in an earlier Virgin Media O2 and Nokia trial in London.
4Deutsche Telekom’s initial laboratory measurements showed a typical improvement in uplink data throughput of between 10% and 15%.

Vodafone Three announced on November 18, 2026, that it was working with Nokia and Nvidia to explore integrating artificial intelligence into its radio access network, with trials planned in the UK. The programme will begin in a controlled laboratory environment, then move to a field trial at a site selected by the company, according to a Vodafone Three statement. The statement did not specify when either phase would start, describing them instead as planned steps in the programme. The company will compare test results with the performance of its existing network to assess whether the new technologies improve capacity, reliability and spectrum efficiency. The partnership is therefore framed as a practical assessment of running network functions on computing infrastructure that also supports AI workloads.

The proposed architecture combines Nokia’s AI-native anyRAN software, Nvidia’s Aerial RAN Computer accelerated computing platform and Red Hat OpenShift, according to the Vodafone Three statement. This setup will allow radio and AI functions to be tested in a shared computing environment, one of the potential applications identified by the company. Possible uses include radio optimisation, network traffic management and improved energy efficiency. The list also includes detecting anomalies, improving coverage and capacity, and assuring quality of service. Vodafone Three is also examining whether accelerated computing resources can be shared between radio access network workloads and AI applications, without presenting these areas as results already achieved.

The trial’s initial use cases include machine-learning-based channel estimation and link adaptation, the company said. These applications relate to estimating signal conditions and adjusting data transmission as radio conditions change; Vodafone Three will test both as part of the programme. The company said comparing results with the existing network would help it identify where AI could deliver the greatest benefit. The announced metrics include capacity, reliability and spectrum efficiency, while the company examines how to scale the capabilities if the tests prove worthwhile. The statement gave no target figures for improvement in these metrics or preliminary trial results.

Vodafone Three links network development to the potential to offer more differentiated connectivity plans and open up new commercial opportunities. The statement cites the existing SuperMobile service as an example of this approach, saying it offers speeds of up to four times those of standard plans in congested areas. The service also guarantees a minimum speed of 15 megabits per second, according to the statement. The company presents the service in a broader commercial context that network capabilities could support, while the benefits of the AI technologies themselves remain subject to testing. The statement does not say whether the new trials will change the existing service or its terms.

The UK trial comes as telecoms companies and equipment suppliers test AI applications in networks. In an earlier live trial at a site in Paddington, London, Virgin Media O2 and Nokia said spectrum efficiency improved by about 20%. Nokia also said in October 2026 that initial laboratory measurements by Deutsche Telekom of AI-based channel estimation showed a typical improvement in uplink data throughput of between 10% and 15%, with greater gains in more complex radio conditions. The work was scheduled to move to field validation. These examples provide context for the UK tests, but relate to other programmes and networks and do not represent results from Vodafone Three’s trial.

At group level, Vodafone opened an AI radio access network test centre in Malaga to compare graphics processing units, central processing units and chips designed specifically for telecommunications. The company said the centre’s results would guide deployment and investment decisions for itself and its partners. As an indication of the expansion of 5G use in the UK, the share of Vodafone’s mobile connections using 5G rose to 23.8% in 2025, from 15.0% in 2024, according to Ofcom data covering the periods ending in March of each year. Ofcom estimated Vodafone’s outdoor 5G coverage in England at 67% with high confidence and 54% with very high confidence in July 2025. These figures provide background on network adoption and coverage; they do not measure the performance of the new trial.

Closing prices in the November 19 session offer a separate indicator of market developments in the data provided: Nvidia closed at $145.98 and Nokia at $7.5. These prices alone do not indicate the partnership announcement’s effect on either share. In Vodafone’s broader programme, the group said it plans to conduct its first laboratory call on a 5G massive-MIMO network in December 2026, followed by field trials; this is a date for Vodafone’s programme, not an announced date for Vodafone Three’s UK trials. Nokia also said BT was preparing for a live-network trial in Birmingham during 2027, as part of its European programmes. Vodafone Three’s announcement did not set a timetable for its UK tests.

Sources

Open on chart