GSMA’s Head of AI Initiatives, Louis Powell, and GSMA Intelligence’s Head of Research, Tim Hatt, explore how surging AI traffic is causing operators to think again about the network’s edge—a topic comprehensively covered in a new GSMA Intelligence report ‘Distributed inference: AI adds a new dimension at the edge’
AI is rapidly reshaping industries, and telecoms is no exception. As networks grow more complex and data demands surge, operators are harnessing AI to boost efficiency, enhance services and unlock new revenue. GSMA Intelligence research indicates that 50% of operators view AI as key to revenue growth1, while NVIDIA’s State of AI in Telecommunications report reveals that 97 percent of telco respondents say they are adopting or assessing AI.2
One of the most impactful trends of AI in telecoms is the distribution of compute beyond datacenters, out to the network edge. As AI workloads grow—particularly AI inference, trained models running on live data—operators are leveraging edge to reduce compute processing times, optimise bandwidth, enhance energy efficiency, and strengthen data sovereignty and network resiliency.
From intelligent network automation to enterprise AI applications, Edge AI is becoming a strategic priority for operators looking to maximise AI’s value. This implies building an AI-optimised edge compute infrastructure that can run AI workloads far more efficiently and cost-effectively than standard compute.
AI Along the Network Edge: Where and Why?
Where Distributed AI Inference Plays a Role
The impact of AI on mobile networks is set to be substantial. AI-driven data traffic is expected to rise significantly, with estimates suggesting a threefold increase in cellular data traffic by 2030 on 2024 levels, and AI potentially contributing an additional 20-80% depending on the scenario.
The surge in traffic will come from AI Direct Traffic, such as cost-saving customer-care chatbots, AI assistants, robotics and security cameras and, secondly, AI Influenced Traffic, or those applications that benefit from AI enhancements, such as video streaming optimisation, social media recommendations and AR-based object recognition.
To address this growth, operators are exploring solutions such as AI-powered network automation, intelligent workload distribution across cloud and edge processing, strategic collaboration with AI Platform Providers and engaging AI Ecosystem players.
So where does AI fit within the telecom network? It can be deployed at multiple points along the edge, each strategically positioned to deliver distinct benefits based on the specific use case. At the device edge, AI runs directly on smartphones, IoT sensors and autonomous systems, enabling real-time decision-making with minimal compute processing times—critical for physical AI applications such as autonomous cars or robotics.
A step further from the user is the telco network, consisting of the network edge and distributed data centers (for core and OSS/BSS). For example, deploying compute at RAN sites allows for intelligent workload distribution balancing processing between the cloud and edge to reduce backhaul congestion and enhance network efficiency. Finally, the enterprise/on-premise data center edge is where AI delivers the most value for industries like manufacturing, retail and healthcare, powering applications like AR/VR, industrial automation, security monitoring, video search and summarisation, AI agents and cashierless stores who prioritise data security and compliance.
Why live at the Edge
Costs savings
Edge AI presents a significant cost-saving opportunity for operators. By processing data closer to the source, operators can reduce reliance on centralised cloud infrastructure, leading to lower bandwidth costs, lower computing costs, better application performance and improved network utilisation. AI inference at the edge optimises network traffic distribution, reducing unnecessary data transfers and minimising backhaul congestion.
This offers significant energy savings by avoiding power associated with backhaul transport and datacenter cooling. We estimate the savings are in the order of 50-60%. So, for example, by retaining 30% of traffic at the edge rather than sending to the cloud, an enterprise could save around 20% on energy usage.
Resilience and Compliance
Nations are now prioritising sovereignty and investing in national in-country infrastructures to ensure resiliency and security, preparing for any external uncertainty. Ensuring data generated in a given territory remains in that territory—within national mobile operator networks and data centers—as opposed to being transferred across international borders facilities regulatory compliance for sensitive data and industries.
As AI powers essential systems like transport, hospitals and smart cities, bringing workloads closer to where they operate is crucial. Edge AI enhances resilience by reducing risks from interference—whether adversarial, accidental, or environmental—ensuring reliable and secure performance for critical infrastructure.
Driving Revenue
As we have already mentioned, half of operators see AI as crucial for driving revenues. Furthermore, our data shows that the priority AI use cases for revenue generation are those where inference should happen at the enterprise edge. This includes robotics, cameras, digital twins, and, to a certain extent, other IoT applications. Imagine industrial automation with advanced robotics, security cameras that can predict incidents and predictive maintenance that keeps everything running smoothly.
In retail, think of AI helping to create cashierless stores and smart inventory management systems. And for AR/VR applications, AI can deliver immersive experiences that make gaming and live events more exciting than ever. In addition, whole new revenue streams are emerging, with operators exploring AI-as-a-service models, offering AI-driven compute power to enterprises as a new business model.
Where next?
In “Distributed inference: AI adds a new dimension at the edge” we introduce a framework for evaluating distributed AI—examining where it can be deployed across the telecom network, the technical, regulatory and environmental advantages, and the business case for Edge AI.
However, the report is just the start. To provide deeper insights into how Edge AI benefits different industries, GSMA Intelligence will soon launch an interactive online calculator in collaboration with NVIDIA and Dell Technologies, enabling enterprises and operators to input their specific use cases and assess the potential gains of running inference workloads at the edge. This tool will offer data-driven projections on cost savings, performance improvements, and energy efficiency—helping stakeholders make informed decisions about AI deployment strategies.
Later this year, the GSMA Foundry, our innovation accelerator, aims to launch two projects collaborating with industry leaders to measure and publish the real impact of Edge Inferencing on critical use cases. We are keen to work with partners across the telecoms ecosystem, operators, vendors, system integrators and startups. If you would like to participate then please email [email protected].
It is an exciting time for telcos, but it’s critical they stay ahead of the curve—with the operators who invest in AI-powered edge computing, automation and monetisation set to gain a significant advantage.
For further insights, please reach out to Tim ([email protected]), or Louis ([email protected]).
Notes
1 https://www.gsmaintelligence.com/research/telco-ai-state-of-the-market-q4-2024
2 https://www.nvidia.com/en-us/lp/industries/telecommunications/state-of-ai-in-telecom-survey-report/