Why network foundation models and AI-RAN won’t save telecom

Telecom vendors and operators are promoting “network foundation models” and AI-RAN as a cure for the industry’s structural problems. The pitch is simple: pour vast amounts of logs, traces and configuration data into huge Transformer models, then use the output to automate and optimize radio access networks at scale. According to this vision, AI will cut costs, improve performance and unlock new growth.

The analysis argues this is wishful thinking. Building and running such models demands expensive compute, deep data engineering work and ongoing tuning that most operators are not set up to handle. The economics are unlikely to improve a sector already squeezed by high capex, low margins and slow revenue growth. AI can help with specific, narrow tasks in the network, but it does not fix broken business models, regulatory pressure or intense competition. The piece frames AI-RAN and network foundation models as an engineering and financial trap if treated as a rescue plan rather than a limited tool.