More AI-RAN shenanigans, following some of the 101 stuff from yesterday about AI for RAN and RAN for AI (or whatever it is called), and how it is all becoming more tangible. In another write-up from Intelligent RAN Forum earlier this week (available here), Rakuten Mobile says it is using its celebrated virtualized open RAN foundation to automate network operations, with more than 20% energy savings from a closed-loop system that has achieved ‘level-four’ validation from TM Forum. It is also working with Intel to put AI inference alongside its vRAN workloads – rather than shipping everything back to a central data center.
Which is a glimpse of the future, really – of RAN for AI, or AI on RAN, or networks for AI (etc etc); using the network’s distributed compute environment for inference at the edge. Meanwhile, T-Mobile is approaching the problem differently in the US, adding AI to its self-organizing network (SON) set-up. (Someone somewhere called AI-RAN ‘SON-on-steroids’ this week, related to coverage of Orange’s AI-RAN work with Nokia.) T-Mobile’s so-called AutoPilot solution uses intent-based automation to adjust neighboring cells when sites fail; its Dynamic CX tech predicts demand and adjusts capacity ahead of major events.
T-Mobile says AutoPilot can make network adjustments in half the time, and its SON solution made 30,000 antenna adjustments during a January winter storm (Fern). Both good cases for intelligence inside automation loops, within human-defined guardrails. Which feels closer to AI-native RAN than the most of the usual hype. See also Optus’ AI-RAN case study with Ericsson from yesterday.
James Blackman
Executive Editor
RCR Wireless News
RCR Top Stories
Open AI RAN: Rakuten Mobile is using its open RAN foundation to advance AI-native automation, improve energy efficiency, and bring AI-RAN inference closer to the network edge. Report from Intelligent Ran Forum.
Agentic Qualcomm: Qualcomm has spent years taking its mobile tech into automotive, IoT, PCs and now data centers; agentic AI is the thread that connects those businesses. Sean Kinney has the story.
China NTN ban: The US has banned Huawei et al from NTN licences and access; 3GPP standards and testing are process; Japan and Europe want their own systems; and Starlink is all over the place, in operators and geo-politics.
Frontier diversity: OpenAI is diversifying its AI chip supply chain, pairing Nvidia with custom Broadcom silicon and AMD accelerators as it separates training from inference and seeks greater efficiency, resilience, and negotiating leverage.
Missing memory: Marvell says AI clusters lack a crucial memory tier between HBM and storage, leaving GPUs idle waiting for data. Its optical shared-memory appliance pools tens of terabytes at near-NUMA latency for faster inference.
AI-Powered Telecom Infrastructure
Supermicro, in collaboration with NVIDIA, delivers AI-powered infrastructure tailored for telcos, enhancing operational efficiency, network management, and customer experiences. Explore now
Beyond the Headlines
AI-RAN gains: AI-RAN is becoming practical with Optus reporting live-network gains in link adaptation, coverage prediction, and compensation. But scaling across multi-vendor networks remains a challenge, notes SK Telecom.
On-device AI: Qualcomm’s Snapdragon 8 Elite Gen 6 comes in standard and Extreme versions, both using 2nm silicon and 5GHz Oryon CPUs. The Extreme adds NPU memory, Matrix Cores and greater headroom for on-device agentic AI.
Networks that pay: Build for the business case, says Mimosa Networks, not for fiber-first dogma. As BEAD economics shift, FWA in unlicensed spectrum offers another route to viable networks, from rural America to urban India.
Cambium WISPs: Cambium’s collapse leaves WISPs with a problem with their sprawling installed bases. Airspan’s partial acquisition offers relief but leaves ePMP and other lines in limbo. The challenge is securing support, spares, and migration.
NPUs for telco AI: SK Telecom is putting Korean-made NPU chips into customer-facing services at meaningful scale. Rebellions says its ATOM-Max NPUs are powering four SK services, processing 14m AI requests and 4bn tokens daily.
What We're Reading
Satellite ban: The Senate has passed a ‘secure space’ act, barring the FCC from approving satellite licenses, market access, or earth-station authorizations for ‘covered list’ entities – Huawei, ZTE etc. Implementation is due within one year.
Network AI: T-Mobile is expanding two AI network automation capabilities nationwide: AutoPilot, which adjusts the network when conditions change, and Dynamic CX, which predicts demand and prepares capacity ahead of major events.
Airport P5G: Ontario International Airport is taking private 4G/5G from Boldyn Networks to support security and perimeter monitoring with dedicated spectrum, edge computing, and private cloud across its 1,700-acre site.
AI skills fund: Verizon is investing $70m in a nationwide AI-skills initiative, combining free training from major technology companies with community-based support for job seekers, educators, small businesses and displaced workers.
Liberty agents: Liberty Global has a deal with AI software firm Sierra to deploy conversational AI agents across its European companies, targeting interactions across chat, voice, and text for 80m connections. It is a three year deal.
