Samsung, SK Telecom team up to develop AI-driven 6G tech.

Samsung Electronics and SK Telecom have signed a memorandum of understanding (MOU) to jointly develop and test 6G technologies, with a focus on AI-based radio access network (AI-RAN) systems.

The partnership will see Samsung Research and SK Telecom’s Network Technology Office lead efforts on AI-based channel estimation, distributed multiple-input multiple-output (MIMO) transmission, AI-RAN schedulers, and new network architectures.

AI-based channel estimation aims to improve real-time signal quality, while distributed MIMO uses multiple antennas to support high-speed data transfer across different environments.

Both companies are also collaborating within the AI-RAN Alliance, where they have proposed an AI-based channel estimation method that was recently approved as a work item.

Samsung and SK Telecom plan to test these technologies using SK Telecom’s network infrastructure.

🔗 Source: Samsung

🧠 Food for thought

Implications, context, and why it matters.

Why the AI-RAN Alliance channel estimation work item matters now

AI-based channel estimation is now an official work item in the AI-RAN Alliance (an industry group focused on applying AI to the radio access network (RAN), the portion of mobile networks connecting devices to cell sites) 1. This sets a path with timelines and milestones for 5G-Advanced (the next major evolution of 5G).
Conventional estimation falls short in low signal-to-noise ratio at cell edges (areas at the outer edge of a cell tower’s coverage) where devices have limited transmission power 1. Lab simulations found 30% better cell edge throughput versus rules-based methods 1.
Work proceeds within the Third Generation Partnership Project (3GPP) under Release 18–19 enablers for AI and machine learning (ML) in RAN 2. Integration could reach 5G-Advanced before 6G work finishes, with field checks expected sooner for proof points 3.

Edge AI suppliers can target operators piloting AI-driven RAN functions

SK Telecom and Japan’s SoftBank have moved AI-RAN testing into the field 4. Suppliers of AI accelerators, MLOps tooling, training data, and test gear at the edge have a window to partner 4. Alliance membership rose from 11 to over 80 in one year 5.
GPU platforms deliver performance but burn power 5. CPU choices with custom silicon offer efficiency plus flexibility 3. That split opens work for vendors with benchmarks, energy tuning tools, or hybrid compute that reduce lock-in.
Marvell integrated AI plus ML accelerators into chipsets used by Nokia and Samsung 4. Open-source tooling supports building ML models for RAN optimization, so third parties in channel estimation, beamforming, or interference mitigation can contend with proprietary stacks 4.