SoftBank, Samsung sign MoU to advance AI-RAN research

SoftBank and Samsung have signed a memorandum of understanding to collaborate on research in AI-RAN, the companies said in October 2025.

The agreement covers joint development in four candidate areas: 6G, AI for RAN, AI and RAN, and SoftBank’s large telecom model.

The companies plan to conduct field trials focusing on the 7 GHz band for 6G, and explore new use cases by combining Samsung’s advanced base station implementation technologies with SoftBank’s expertise in network deployment.

They will also research how AI can improve efficiency at the physical layer of radio access networks, and define an AI-RAN orchestrator for integrating AI and RAN workloads on shared infrastructure.

The collaboration will leverage SoftBank’s large telecom model, a generative AI foundation model for telecommunications, to transform network optimization workflows with LLM-based AI, improve performance, and ensure secure data interoperability.

Both companies said they aim to enhance the efficiency and performance of communication networks, and contribute to the commercial deployment of next-generation infrastructure.

🔗 Source: SoftBank

🧠 Food for thought

Implications, context, and why it matters.

The 7 GHz band for 6G remains contested in global standards bodies

SoftBank and Samsung’s 7 GHz field trials target a candidate band for mobile use, yet regulators have not aligned worldwide.
The World Radiocommunication Conference and the International Telecommunication Union have not set global 6G spectrum rules. WRC‑23, the ITU’s treaty‑level spectrum conference, set region‑specific outcomes for 6.425–7.125 GHz. WRC‑27 will study 7.125–8.4 GHz, so 7–8 GHz stays unsettled 1.
Unclear timelines and mismatched allocations could push operators to invest in tech that needs rework if early 6G rolls out on other bands after 2030 2.

LLM vendors can target telcos piloting network optimization AI

SoftBank’s Large Telecom Model hints that carriers are testing LLMs for network tuning, which opens room for specialized AI infrastructure providers.
AT&T and Deutsche Telekom are building AI agents (autonomous software assistants for operational tasks) for RAN operations 3. Separately, 25% of enterprises using generative AI are projected to deploy AI agents by 2025–2026 3.
Third‑party vendors can offer LLMOps (operations tooling for deploying and maintaining LLMs) or secure data interoperability tools. Domain‑specific training datasets add value. These partners can work with telecom operators running AI‑RAN pilots, particularly those without in‑house LLM development capabilities.