AI is increasingly moving from the applications and data centers that power it into the networks that connect them. A new trial by SoftBank and Ericsson in Japan offers a concrete example, with AI being applied directly to a commercial 5G radio access network in real time.
The companies said their validation of Ericsson’s AI-native Scheduler for Link Adaptation delivered gains of up to approximately 25% in spectral efficiency and approximately 50% in downlink user throughput, compared with conventional technology. Across the evaluated locations, both metrics improved by approximately 10% on average.
The significance goes beyond the performance figures. Rather than using AI simply to analyze network data or support operations, the trial applied an AI model directly within the RAN, allowing it to respond to changing radio conditions and optimize link parameters in real time.
That approach could become increasingly relevant as mobile networks face new traffic patterns driven by generative AI, autonomous agents, and other emerging applications. The challenge is no longer simply adding capacity, but making better use of existing network resources as demand becomes more variable and complex.
The SoftBank-Ericsson trial also illustrates a broader point about the AI infrastructure race: the technology stack is changing rapidly, often in places that were previously considered relatively mature. From data center optics and AI computing to antennas and now the RAN itself, established infrastructure is being redesigned around AI.
That raises a bigger question about how operators and infrastructure providers should plan when the pace of technological change itself is becoming harder to predict. In today’s newsletter, Nokia’s Mike Bushong explores that question in “Planning for next: Black swans during an AI frenzy,” examining how the intensity of the AI investment cycle could itself create new sources of disruption.
Juan Pedro Tomas
Editor
RCR Wireless News
RCR Top Stories
AI reshapes test demand: Keysight’s AI-driven wireline growth is increasing testing demand across R&D and manufacturing, while rising technology complexity, supply constraints, and emerging 6G applications expand its opportunity.
KT builds sovereign AI: KT has unveiled an on-premise sovereign AI server combining Korean NPU silicon, an LLM and software platform, targeting organizations that need generative AI without external cloud connectivity.
AI’s black swan: Nokia’s Mike Bushong argues that AI’s unprecedented investment, technology shifts and supply constraints could create the conditions for the next black swan, requiring more flexible corporate planning.
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
Nvidia Feynman GPU: Nvidia is reportedly designing its next-gen “Feynman” GPUs to debut on TSMC’s A16 node – a 1.6nm-class process – rather than the standard 2nm (N2) generation. That would make Nvidia the first major customer for A16.
Power limits AI buildout: Dell’Oro’s Baron Fung says AI infrastructure could account for roughly 40% of $3 trillion in data center capex by 2030, but power availability remains the biggest constraint.
Telecom AI ambitions: Telecom leaders call AI a key revenue driver, yet only a quarter of operators are ready to scale it. HCLTech’s survey exposes the gap in between.
DC scenarios: NTT Global Data Centers models three AI data center growth scenarios through 2030, concluding that power, supply chains, and infrastructure constraints—not demand—will determine whether AI capacity can keep pace.
10 years for HCF: HKT’s hollow-core fiber AI ‘superhighway’ highlights the tech’s promise for ultra low-latency DCI networks, but analyst CRU Group says costs, ecosystems, and production mean at-scale adoption is years away.
What We're Reading
Charter absorbs Cox: Charter has completed its Cox acquisition, expanding Spectrum across 45 states and bringing broadband, mobile and entertainment services to new markets while promising greater scale, investment and customer benefits.
Unicom shifts compute: China Unicom’s first-half profit fell sharply despite stronger cash flow, as compute revenue grew and investment surged more than 80%, underscoring the operator’s accelerating shift toward AI infrastructure.
EE prioritizes traffic: EE launches Fast Lane, a commercial 5G+ network slice that prioritizes consumer and small-business connectivity during congestion, targeting smoother streaming, payments, and critical services at busy locations.
du trains SMEs: du Business is launching UAE-wide training for 10,000 SMEs on Agentic AI, emphasizing practical workshops and business applications aimed at improving efficiency, customer engagement, and digital competitiveness.
O2 expands 5G+: O2 has switched on its standalone 5G+ network across Kent, extending faster speeds, improved reliability and lower latency to up to 1.6 million residents as part of its transformation plan.
