The press release read like a roster rather than an announcement, and that was the point. On September 16, Nokia named eight telecom operators across four continents that are moving its AI-powered radio access network, or AI-RAN, from early evaluation into laboratory testing and live trials. The list stretched from A1 Group in Europe to Chunghwa Telecom, du, e&, Mobily, stc, TPG Telecom, and Zain in Saudi Arabia and the wider region.
The geography is the message. Nokia has spent years trying to convince carriers that AI belongs inside the radio network itself, not just in the data centers that sit behind it, and the fact that operators in North America, Europe, Asia-Pacific, and the Middle East are now running tests suggests the argument is landing in more than one market.
Radio access networks are the least glamorous part of the telecom stack and the most expensive. They are the base stations and antennas that connect phones to the network, and carriers spend more on them than on almost anything else. If AI can squeeze more capacity out of the same equipment, the savings run into the billions, which is why every major equipment vendor is now chasing the same story.
Nokia said its previously announced test program with NTT DOCOMO is also moving forward, and that it has already launched a commercial AI-native RAN platform. The company named T-Mobile, SoftBank, and Indosat Ooredoo Hutchison as partners in evaluations, a set of names that gives the effort scale across three very different markets.
The relationship with Nvidia is the second thread running through the announcement. Nokia said it is deepening its collaboration with the chipmaker on the AI-RAN platform and on software development tools. Nvidia has been pushing hard into the telecom equipment market, arguing that its GPUs can run AI workloads directly at the edge of the network, and Nokia has become one of its most visible allies in that campaign.
The logic for Nvidia is straightforward. Its data-center business is enormous, but the telecom market represents a new frontier where AI compute would sit physically close to the customer. If carriers start putting GPUs into base stations to run AI services locally, Nvidia sells a new category of chips, and Nokia sells the equipment they plug into.
For Nokia, the bet is about avoiding a repeat of history. The company was slow to the early cloud buildout and lost ground to rivals in the software-defined networking shift. AI-RAN is the next transition, and Nokia is moving earlier this time, signing up operators and a chip partner before the market has fully decided what AI in the network is even for.
The use cases are still taking shape. Operators talk about using AI to manage spectrum in real time, to predict and prevent congestion, and eventually to host AI applications for enterprises at the network edge. None of it is proven at scale, which is exactly why the current phase is tests and trials rather than orders.
Analysts cautioned that trials do not become revenue quickly. Telecom operators are deliberate buyers, and a live trial is a long way from a commercial deployment across a national network. But the breadth of the trial list matters, because it signals that carriers in very different regulatory and competitive environments are asking the same question about whether AI can cut their costs.
Nokia framed the announcement as momentum rather than a destination. The company said AI-RAN is accelerating globally, and the evidence is the map of operators willing to spend money testing it. Whether that turns into a durable market, and whether Nokia or a rival captures it, is the question the next year of trials is supposed to answer.
A radio access network is the part of the network that actually reaches the customer, the antennas and base stations that carry the signal the last mile. It is the largest single item in most carriers’ capital budgets, and it has traditionally been the most conservative part of the network to change. Nokia’s argument is that AI can change that, by managing spectrum and traffic in real time and by turning the edge of the network into a place where AI workloads can run.
The competitive field is crowded. Ericsson is pursuing the same transition, and the Chinese vendors have their own AI-RAN programs, though their access to Western markets is restricted. Nokia’s partnership with Nvidia gives it a chip ally that the others do not have in the same way, and the trial list it published is a bid to show that the technology is moving from slides to field tests faster than its rivals can match.
The stakes go beyond one vendor’s share. If AI-RAN works, the network edge becomes a computing platform, and the companies that own it will collect a toll on every AI application that runs there. That is why the trials matter to the operators as much as to the vendors, and why eight names across four continents constitute more than a publicity list.
