Telecom Giants Launch 'Open Telco AI'

The "Open Telco AI" initiative, launched at MWC Barcelona 2026, directly addresses the significant performance gap of general-purpose AI in the telecom sector. Research from GSMA Intelligence revealed that only 16% of generative AI deployments in the industry have been applied to network operations, the industry's largest cost center, because today's frontier AI models struggle with the sector's specialized language and data. This industry-wide collaboration aims to create AI that "speaks telco," a capability currently lacking in even the most advanced models which fail to accurately interpret complex network data or understand technical standards documentation. The goal is to move beyond generic tools that can produce inaccurate or impractical recommendations, a critical issue in an industry demanding high precision, reliability, and safety. As a founding supporter, AT&T is contributing a family of open telco-specific AI models. These models are trained on publicly available data and are designed to be hardware and cloud-agnostic, allowing for broad application across projects of varying scales. This initiative provides a portal for these models, along with curated datasets and tools, to foster development. Chipmaker AMD, another founding partner, is providing the essential computing power for training, fine-tuning, and evaluating these specialized AI models. AMD's GPU platforms and open toolchains will be accessible to the initiative's participants, including cloud partner TensorWave, removing significant infrastructure cost barriers for development and testing. Progress and model performance will be measured by the "Telco Capability Index," a public leaderboard that assesses AI models against seven specific telecom benchmarks. This allows for transparent evaluation of how well different models perform on critical, real-world industry tasks. The initiative is backed by a wide array of over 46 organizations, including major operators like China Mobile, Deutsche Telekom, Orange, Telefónica, and Vodafone. Technology providers such as Google Cloud, IBM, and Nvidia are also contributing, alongside numerous academic institutions, highlighting a broad consensus on the need for a collaborative approach to building domain-specific AI. This collaborative effort extends to community programs designed to solve real-world problems. For instance, the AI Telco Troubleshooting Challenge has already attracted over 1,000 participants, fostering a community of developers and researchers focused on resolving genuine telecom AI issues. The long-term vision is to create a more efficient, automated, and intelligent network ecosystem. By developing AI that understands the unique complexities of telecommunications, the industry aims to enhance customer experience, reduce operational friction, and unlock new value and services.