According to GSMA Intelligence, 2025 is the year telcos move from talk to action on artificial intelligence.
The report “Telco AI: State of the Market, Q2 2025” shows the industry shifting from strategy-setting to practical implementation.
Where 2024 was about building the strategic case for AI, 2025 is focused on measuring progress and delivering tangible results.
First, the easy wins
Operators are prioritizing low-risk, high-impact deployments, with the clearest results appearing in customer support. AI chatbots and automated call-centre systems account for about 47% of tracked telco AI projects.
The primary objective is cost reduction through automation. Network operations follow, representing roughly 20% of deployments and covering tasks such as predictive maintenance and automated fault detection to keep services running reliably.
Overall, an estimated 80–85% of telco AI initiatives aim to lower internal costs and boost efficiency. Driving new revenue currently makes up a smaller portion—around 10–20% of projects.
These efforts are not limited to pilots. Approximately 60% of telco AI systems are already live and integrated into daily operations, while the remaining 40% are in trial or planning phases. GSMA Intelligence interprets this as the start of a continuous cycle of experimenting, deploying, and refining AI-driven solutions.
Edge computing is the next frontier for telco AI
Although near-term priorities emphasize cost savings, the more transformative potential lies at the network edge. Telcos are positioning edge computing as the platform for future services and new revenue streams.
Model training remains concentrated in large cloud data centres, but real-time use of trained models—known as inference—is increasingly moving closer to users at the edge. Running inference at the edge can be 30–40% less expensive, reduces latency, and helps keep sensitive data local for improved privacy and security.
This shift enables new business-focused services, with early use cases including industrial robotics, intelligent video analytics for security, and sophisticated digital twins for manufacturing and logistics.
The report stresses the main challenge is commercial rather than technical: operators must find viable business models in a market that is rapidly becoming crowded with competitors and specialised providers.
A new cast of AI partners for telcos
Telcos are relying on a broad ecosystem of technology partners as they build AI capabilities. In the GPU market, Nvidia holds a clear first-mover advantage, supplying the specialised hardware needed for large-scale model training and inference.
Meanwhile, newer entrants such as Perplexity have emerged as noteworthy partners, particularly for AI-powered search and answer capabilities. Some operators see collaborating with alternative AI search providers as a way to diversify dependencies and avoid overreliance on any single dominant supplier.
Preparing the workforce is another key priority. Around 63% of operators cite upskilling staff as essential to manage and govern AI systems effectively, reflecting a widespread need for in-house expertise.
For now, telcos are using AI primarily as a tool to address pressing cost and efficiency challenges. Simultaneously, they are investing at the edge and forging partnerships to enable future revenue-generating services. The coming years are likely to see ongoing experimentation and refinement as operators work to unlock AI’s full commercial potential.
(Photo by Leon Kohle)
See also: SK Telecom releases compact AI model for mobile use on Hugging Face
Want to learn more about AI and big data from industry leaders? Consider attending AI & Big Data Expo events held in Amsterdam, California, and London. These industry gatherings bring together leaders and practitioners to explore AI, data platforms, and enterprise applications across multiple sectors.
Explore other upcoming enterprise technology events and webinars powered by TechForge.