The GSMA recently unveiled the Open Telco AI initiative at the Mobile World Congress 2026 in Barcelona. This global initiative aims to address the specific needs of the telecom industry by developing telco-grade artificial intelligence through open collaboration. The traditional AI models fall short in precision and reliability for telecom network operations, which are highly demanding and regulated environments.
Louis Powell, Director of AI Initiatives at GSMA, articulated the issue succinctly, noting that current AI models “do not yet speak telco.” The GSMA seeks to convene operators, vendors, AI developers, and academics to create shared models, data, and computing resources tailored for telecom. Such collaboration is crucial, given that only 16% of telecom GenAI implementations currently affect network operations.
Highlighting this consortium’s ambition, Johan Ottosson, VP Strategy at Arelion, commented, “The telco world is notoriously slow, and such federations have been launched – and fallen apart – at previous occasions.” However, he suggests the initiative’s potential is significant if it learns from past efforts and starts small, generating tangible progress that paves the way for broader standardization.
The initiative’s founding partners, AT&T and AMD, provide complementary technical resources. AT&T offers open-source models trained on public data, ensuring hardware and cloud neutrality. This approach underscores that AI can deliver impactful results across various computing scales. Meanwhile, AMD provides computing capacity through its partner, TensorWave, offering participants necessary infrastructure without vendor lock-in.
The technical ecosystem supporting this initiative includes advanced models like RFGPT from Khalifa University and AdaptKey AI’s Large Telco Model, built on Nvidia Nemotron. These open datasets from leading universities enhance the initiative’s collaborative foundation. Ottosson points out, “If a consortium can federate this, it solves a unique problem that a large budget just can’t solve with sheer money.”
Benchmarking is also a key focus. The Telco Capability Index evaluates model performance based on telecom-specific tasks. This framework helps in distinguishing genuine progress from mere noise, making improvements measurable. Moreover, community engagement is robust, with challenges like the “AI Telco Troubleshooting Challenge” attracting over 1,000 registrations. These events help test open models and tools in simulated real-world conditions.
The breadth of partners, including Google Cloud, Deutsche Telekom, China Mobile, Vodafone, Telefónica, and IBM, adds credibility. Ottosson suggests focusing on augmenting humans to improve response times as a first step, which could pave the way for fully autonomous solutions in the future. The sustained involvement of these industry giants will be crucial in ensuring the initiative meets its ambitious goals.
