From Tokyo Streets to Smarter Cells: Samsung and KDDI Deliver AI-Driven RAN Gains on Live 5G

In the bustling districts of Tokyo, where commuters stream through dense urban canyons and suburban towers hum with evening data surges, a quiet revolution unfolded over several months starting in late 2025. Engineers at KDDI monitored dashboards as Samsung’s AI-powered RAN Speed Optimizer (RSO) quietly retuned hundreds of cells across 100 MHz of 3.7 GHz TDD spectrum. What began as a field trial on a live 5G Standalone network evolved into tangible proof that artificial intelligence could deliver precise, cell-by-cell optimization at commercial scale.

The results, announced in early July 2026, speak volumes: an average 31% increase in downlink throughput across the trial area, with peaks of 52% in the most crowded urban zones. This wasn’t lab magic or simulated traffic—it happened amid real-world variability in dense cities, suburbs, and rural fringes around Tokyo.

The Trial That Proved Per-Cell Intelligence Works

Traditional RAN optimization has long relied on cluster-level parameters—applying the same settings across groups of cells, a blunt instrument that often left performance on the table in heterogeneous environments. Samsung’s RSO changes that paradigm. Its AI-based prediction model ingests site-specific environment data, automatically recommends and applies optimized parameters tailored to each individual cell.

Conducted across hundreds of cells in varied conditions, the trial trained robust AI models on actual traffic patterns. Peak-hour gains demonstrated the system’s ability to adapt dynamically, reducing manual intervention and unlocking efficiency. For operators, this translates to lower operational costs; for users, faster, more reliable connections whether streaming, browsing, or connecting IoT devices.

KDDI’s Chief Network Officer Kazuhiro Furuhata highlighted the milestone: “Combining KDDI’s accumulated expertise in network innovation and Samsung’s technical leadership, this field trial proves that individual tuning for cells—a long-standing industry challenge—has now become a reality through the integration of AI.”

Samsung’s June Moon, Executive Vice President and Head of R&D for Networks Business, echoed the progress: “Since 2024, we have been actively testing and training our AI-powered RSO technology in the field… This trial with KDDI exemplified how Samsung’s AI-powered innovation can bring advanced optimization to live commercial networks while facilitating steady, seamless connectivity in an array of different network environments.”

Building on Virtualized Foundations Toward AI-Native Operations

The success rests on KDDI and Samsung’s longstanding collaboration in fully virtualized network deployments. This vRAN base provides the disaggregated, cloud-native architecture essential for layering advanced AI capabilities. As networks evolve, these foundations accelerate the shift to AI-native operations, where intelligence is embedded deeper into the RAN fabric.

Samsung positions RSO within its broader CognitiV Network Operations Suite (NOS)—a portfolio of AI-powered automation solutions, agents, and intelligent applications. The trial’s outcomes validate the move beyond conventional approaches, setting the stage for wider commercial rollout of similar AI-driven tools.

Why This Matters for Open RAN and the AI-RAN Future

While the announcement centers on 5G SA optimization, it carries significant implications for the Open RAN ecosystem. Samsung has been a vocal proponent of vRAN, Open RAN, and AI-RAN pathways. The ability to achieve granular, AI-tuned performance on live networks underscores how open interfaces and virtualization enable faster integration of intelligent applications—precisely the promise of RAN Intelligent Controllers (RICs) and xApps/rApps in O-RAN architectures.

Operators embracing Open RAN principles often report greater optimism about adopting AI innovations at the edge. This trial illustrates that momentum: AI models trained on real traffic can optimize parameters in ways that static or cluster-based methods cannot, paving the way for more autonomous, intent-driven networks heading into 6G.

The companies plan to extend evaluations of AI-based optimization for broader applications, leveraging the virtualized infrastructure as a springboard. This aligns with industry-wide momentum toward AI-RAN, where compute resources are shared across narrow controllers and intelligence becomes native rather than bolted on.

Broader Industry Context and Next Steps

As 2026 unfolds, field validations like this move AI-RAN discussions from roadmaps to reality. They complement ongoing standards work in bodies like the O-RAN Alliance and AI-RAN Alliance, while highlighting practical gains in throughput, efficiency, and reduced human oversight.

For the telecom audience, the takeaway is clear: AI-driven per-cell optimization is no longer theoretical. It is delivering measurable results today on commercial infrastructure, with Open RAN and virtualization serving as the critical enablers for scaling these capabilities.

Samsung and KDDI’s collaboration continues, focused on pushing AI boundaries to elevate customer experience. In a world of exploding data demands and spectrum constraints, such advances signal a future where networks don’t just connect—they intelligently anticipate and adapt.

Sources

• https://telecomdrive.com/samsung-kddi-validate-ai-powered-ran-optimisation-on-live-5g-sa/
• https://news.samsung.com/ (referenced in the article as primary source)