The Case for Agentic Base Stations in 6G RAN

As we head toward 6G, the industry faces a fundamental question: how can we integrate artificial intelligence into the radio access network (RAN) in a way that is both effective and scalable? Despite years of hype about AI in telecoms, much of the discussion has remained abstract, full of promises of “AI-driven growth” and “enhanced services”, but falling short on concrete, implementable strategies.

Beyond the AI Hype: Toward AI-Native RAN

In 5G Advanced, the RAN has taken its first steps toward AI integration, as reflected in ongoing work by 3GPP. However, these have primarily been bolt-on techniques, not a rethinking of the base station architecture itself. To realise the vision of an AI-native 6G network, we need a fundamental shift – a move from discrete “AI-enhanced” functions to fully agentic systems.

An agent, in AI terms, is considerably more than just a rules engine or inference model. According to DeepMind and classic cybernetics theory, it’s a system that observes its environment, builds a model of how it works, and adapts its behaviour accordingly. Powered by these world models, agentic base stations wouldn’t execute pre-trained models but instead learn locally, adapt autonomously, and collaborate intelligently with neighbouring nodes.

The Pitfalls of Centralised AI

The default approach to AI for the RAN relies on placing AI in the cloud, either for both training and inference, or at the very least, for offline training of models. Every base station location is different and has dynamic characteristics over multiple timescales (for RF environment and traffic). Both of those factors make remote centralised training an unfavourable approach due to the inherent lag in train/deployment cycle. This approach struggles to meet the needs of effective RAN operation while also introducing additional CAPEX and OPEX burdens, making it impractical for large-scale deployment.

Demands on Base Station AI

Baseband processing at the RAN edge faces stringent technical constraints, including ultra-low latency, tight synchronisation, and high power efficiency. Offloading AI workloads to general-purpose cloud compute, or even to air-cooled edge servers, is impractical at scale. Instead, baseband AI must be tightly co-designed with its hardware environment, integrating DSP acceleration, dedicated inference engines, and high-throughput I/O to meet wireless telecom-grade performance requirements.

What Makes a Base Station Agentic?

An agentic base station is a self-regulating node that uses AI not only for optimising PHY layer operations (like beamforming or interference mitigation) but also for making autonomous decisions across L1, L2, and L3. Crucially, these agents would learn from local data, reflecting their specific RF environment, traffic patterns, and user behaviour rather than relying solely on centralised training sets. This approach also means that agentic base stations can work together in multi-agent configurations, sharing knowledge and optimising jointly for system-wide goals.

Making AI-Native RAN a Reality

Despite early application, cloud-first strategies and repurposed data centre AI chips will not fully drive the transition to AI-native RAN. It requires bespoke, embedded AI architectures tailored to the unique constraints of modern wireless telecoms. An approach that involves rethinking silicon, software, and system design, with a focus on deterministic performance and autonomous adaptability.

RANsemi’s vision for agentic base stations offers a compelling roadmap: tightly integrated AI and baseband processing, local learning at the edge, and a shift from passive execution to proactive optimisation. In the world of 6G, the base station must do more than transmit and receive – it must understand, decide, and act.

Only truly agentic base stations can scale AI in the RAN and close the gap between theoretical potential and real-world performance.

Shaping the Architecture of AI-native RAN Starts Now

If you’re exploring what 6G will demand from RAN base station design, get in touch with our team.

Author: Oliver Davies, VP Marketing, RANsemi