Work starts with Nokia and NVIDIA to test AI-powered networks
Vodafone today opened a European testing hub in Málaga, Spain, to assess how AI in mobile Radio Access Networks (RAN) can improve connectivity for customers and turn the network into a platform for new AI service delivery.
The opening coincides with today’s announcement that VodafoneThree is working with Nokia and NVIDIA to explore AI-powered mobile networks through planned UK lab and field trials (see separate press release).
Vodafone’s Málaga AI-RAN lab also brings together Nokia, NVIDIA, academic partners, other equipment vendors and software developers to test the latest chips, including GPUs and CPUs, for shared RAN and AI workloads. The findings will guide the AI-RAN deployment decisions of Vodafone and its partners to maximise the value of future investment.
AI embedded in everyday applications
The opening of Vodafone’s AI-RAN lab comes as AI is increasingly embedded in everyday applications, connected devices and digital services, which will fundamentally change network traffic patterns over time. Radio access networks connect millions of smartphones and other devices wirelessly to mobile networks. Their coverage and proximity to users make them essential for carrying AI traffic and a key focus for telecoms and technology companies.
Kyriakos Exadaktylos, Head of RAN Evolution in Group Technology, said: “The Málaga Lab uses our strong foundation in Open RAN to create a multi-vendor AI-RAN innovation and validation hub. By bringing together leading RAN, compute and AI technologies, we will establish independent, repeatable benchmarks that show where AI delivers real value in mobile networks.
“It will demonstrate how AI can improve network performance and energy efficiency, how AI and RAN workloads can share computing infrastructure securely and efficiently, and how intelligence at the network edge can enable new customer services. Málaga will help us turn AI-RAN ambitions into measurable results and lay the foundations for 6G.”
Vodafone’s testing environment aims to support this broader ecosystem of AI services using three approaches: AI-for-RAN, AI-and-RAN and AI-on-RAN. Drawing on its Open RAN expertise and tools for combining different hardware and software, Vodafone is offering a multi-platform test bed that brings together key industry players to help strengthen Europe’s global role in AI development and standardisation.
AI-for-RAN: improving the customer connection
AI-for-RAN uses AI to enhance and optimize network performance and efficiency. Vodafone will evaluate features like beamforming that direct radio signals towards users, the use of AI to predict changing radio conditions and allocate network resources, as well as tools to optimise energy use. Customers benefit from more capacity and better connections, particularly when networks are busy. It also paves the way for differentiated connectivity propositions, building on existing services like Vodafone SuperMobileTM, which offers customers in the UK speeds up to four times faster than standard plans in busy areas and guaranteed minimum speeds of 15 Mbps.
The lab will validate AI enabled advanced radio algorithms for channel estimation, link adaptation, beamforming, multi-user MIMO as well as network optimisation software, including applications known as rApps, xApps and dAPPs, from equipment vendors, third-party developers and Vodafone’s own teams. One focus will be on improving spectral efficiency, which involves carrying more data over the same radio frequencies. This is in line with Nokia’s ambition to double spectral efficiency by 2028.
AI-and-RAN: testing a new computing foundation
AI-and-RAN explores whether different mobile network functions and AI applications can run on shared computing infrastructure. Vodafone is working with Nokia and NVIDIA to test accelerated computing systems alongside purpose-built telecoms chips, known as ASICs, and the x86 processors used in Open RAN. The Málaga lab is equipped with QCT servers hosting the NVIDIA Aerial RAN Computer MGX 2U Short platform, and Intel Xeon 6 Granite Rapids servers to assess these workloads.
Sharing computing resources means that the same infrastructure can support both enhanced connectivity and AI services. Vodafone plans to assess silicon performance later this year using industry-wide Standard Performance Evaluation Corporation (SPEC) benchmarks. It will assess AI performance using MLPerf (developed by a consortium of AI leaders from academia, research labs, and industry) and RAN performance benchmarking through the AI-RAN Alliance. The comparisons are intended to inform architecture and investment decisions and not assume one type of chipset is best for every task.
Vodafone plans to make its first 5G massive-MIMO (a method to increase capacity without additional spectrum) laboratory-based call in December 2026, followed by field trials. The initial focus is on uplink performance (how effectively devices send data such as user-generated videos to the network), as well as AI’s ability to improve operations and lifecycle management, while reducing total cost of ownership.
Further work with NVIDIA will examine radio and distributed processing architectures, the evolution of network chips and the use of AI. Vodafone intends to explore the NVIDIA Aerial portfolio and associated tools for designing, simulating and operating wireless networks.
AI-on-RAN: bringing processing closer to users
AI-on-RAN examines how distributed mobile network infrastructure could host AI applications near end users such as for Physical AI, vision AI, personal AI, transportation AI and robotics. Processing data closer to its source promises to reduce response times required to effectively run robotics, computer vision and extended-reality services. Vodafone's planned assessments include integrated sensing and communications (ISAC), which combines wireless connectivity with sensing capabilities, for example, to detect intruders at sensitive sites like airports.
Together, the three AI-RAN programmes will give developers a foundation for bringing AI services to consumers, businesses and public-sector organisations, including better connections to carry their data, shared infrastructure to run workloads and processing closer to users. The planned lab work and field trials this year and next will test whether the technology can deliver customer benefits and a credible business case before wider deployment.
Want to know more?
- Read our interview with Nadia Benabdallah on building the network of the future.
- Find out how we’re creating the network of tomorrow.
- Hear from Iain Wilkinson about what’s next for our networks in Ask the Experts.