Mobily and Ericsson target Level 4 autonomy for Saudi networks

Mobily and Ericsson signed a memorandum of understanding at LEAP 2026 in Riyadh to apply artificial intelligence inside the radio access network, using software components called rApps running on Ericsson's Intelligent Automation Platform, and to advance Mobily towards Level 4 network autonomy in selected use cases. The two things are halves of one agreement rather than separate programmes, and the connection between them is the point. The AI work is the mechanism; Level 4 autonomy is what it is supposed to achieve. The radio access network, or RAN, is the part of a mobile network between the handset and the core: the antennas, radios and baseband equipment that turn a phone's signal into traffic the network can route. It is the most expensive part of a mobile operator's estate to build and by some distance the most complex to run. A national RAN is tens of thousands of cells, each with hundreds of configurable parameters, interacting with its neighbours, in conditions that change by the hour as people move around a city. Operators have tuned that estate with rules and human judgement for thirty years, and the approach has reached its limit. There are too many parameters, changing too fast, for engineers to optimise by hand. The case for putting machine learning into the RAN is that the problem is now genuinely beyond manual tractability. The agreement covers AI functions running in two places at once. Some sit close to the radio itself — in RAN compute, in the radios and in the active antenna systems — where decisions have to be made in fractions of a second: how to shape a beam, how to schedule a transmission. Others run centrally as rApps on the Ericsson Intelligent Automation Platform, where a system can see the whole network at once and reason about patterns no individual cell can perceive. An rApp is best understood as an application for the network rather than for a user: a self-contained piece of software that consumes network data and issues optimisation decisions back. Mobily will join the platform's ecosystem, giving it access to rApps built by other vendors and developers rather than only those Ericsson writes. Level 4 is a rung on an industry scale for network autonomy that runs from zero, where everything is manual, to five, where the network runs itself entirely. At Level 4, the network takes over the closed loop for defined domains: it monitors its own condition, decides what to change, makes the change, and verifies the result, with people setting intent and objectives rather than approving individual actions. The distinction from ordinary network monitoring is total. Monitoring tells an engineer something is wrong. Level 4 means the network noticed, diagnosed, corrected and confirmed — and the engineer reads about it afterwards. That is a demanding standard, and the industry has been more enthusiastic about the label than about reaching it. It is also why the wording of this agreement matters: it targets Level 4 in selected use cases, not across the network. Nobody is running a national mobile network at full Level 4 autonomy, and the honest version of the ambition is that particular functions — energy optimisation, some classes of fault recovery, certain performance tuning loops — run closed-loop while everything else stays supervised. Mobily and Ericsson have a demonstrated basis for the claim. The two companies showed Level 4 automation capability during Hajj 2025, which is close to an ideal proving ground: a temporary, enormous, geographically concentrated surge in demand, with no tolerance for failure and no possibility of asking the users to come back later. A network that can reconfigure itself under those conditions has been tested in a way a lab cannot replicate. Energy is the practical driver behind much of this. Radio access networks consume the majority of a mobile operator's electricity, and a large share of it powers capacity that is not being used — cells lit up through the small hours for traffic that is not there. Software that can shut down and restore capacity safely, in response to real demand, produces savings that are measurable on a bill. It is also the kind of decision an operator is willing to let a machine make, because the failure mode is a brief drop in capacity rather than an outage. Mobily is now running automation programmes with both of the major European vendors, having signed an MoU with Nokia on AI-powered network efficiency at this same event. Whether that reflects a deliberate multi-vendor strategy or two parallel explorations is not something either agreement makes clear. Ericsson has built its LEAP presence around 5G Advanced and AI networks , and Nokia has just opened a Riyadh R&D centre working on precisely this class of software — including rApps and self-organising networks. What has not been published is any of the commercial or operational specifics: which use cases go first, over what timeframe, across how much of Mobily's footprint, or wha