31 October 2025
Digital Twins & AI: Moving Beyond the Hype in Mobile Networks
Talk of AI and digital twins has been inescapable at industry events in recent years. In an industry with increasingly tight margins, operators are looking for new ways to improve the efficiency of capital spend whilst maximising customer experience. Can AI and digital twins deliver tangible value to mobile network operators faced with this challenge?
What is a digital twin?
A digital twin is a living model of how a (mobile) network is built and how it behaves. It mirrors radio sites, sectors, spectrum layers and policies, but also the messier realities that shape performance: traffic patterns, interference and the built environment. Crucially, it stays aligned to the real network with live (or near-live) data, so engineering and optimisation teams do not have to plan against last month’s picture.
This makes a twin different to legacy planning tools. Rather than relying on static lab tests, engineers can use the twin to see the current network in context and try what-ifs before committing to new investment decisions. This includes adding (or removing) 4G & 5G carriers, retuning sectors by changing antenna tilts or power, or re-routing traffic and then quantifying the impact on coverage, throughput, latency, energy and cost before anything goes live.
The role of AI in furthering the use of digital twins
The use of AI has the potential to further enhance digital twins from being exciting near-live network models to value-adding decision tools.
AI allows efficient sifting through noisy data (telemetry) to spot emerging issues and demand spikes and further compares thousands of safe alternative deployment options in the twin to identify the best options. GenAI (AI that creates new content on demand) can lower the friction further by auto-generating synthetic data1 to cover rare events and simplifying model updates (e.g. mapping messy data into the twin). Natural-language copilots also allow engineers to actively query the twin on live network KPIs, and develop code that optimises RAN performance and deployment costs in hours rather than weeks2.
What benefits can operators expect from digital twins & AI?
Whilst the deployment of digital twins is still in the early stages, some mobile operators are already reporting tangible benefits being delivered from initial implementations:
- Improved capacity and user experience: In Taiwan, Chunghwa Telecom reported a 14% capacity improvement and better user experience during the usual traffic peak at New Year’s Eve3, whilst Telefónica Germany reports a reduction in occurrences of high-loading sites with significant customer impacts by 90%4.
- Cost savings from optimised energy use: Verizon claims to have achieved more than USD100 million of annual energy savings from AI-driven consumption models within a digital twin5.
- Time savings and transport layer improvements: Telefónica Germany reports 80% time savings on analyses across its planning, operations and optimisation functions and 40% fewer transport layer capacity issues after mapping its 28 000-site network and transport routes into its own digital twin model.
We expect these developments to be the first step towards increased automation. In particular, we would expect the capex efficiency of operators to improve significantly in the coming years, where improvements in customer experience should go hand-in-hand with reduced spending, as operators get more ‘bang-for-the-buck’ on any investments made.
With all this in mind, we understand that there are still several obstacles to be overcome:
- Data plumbing. Clean, timely, joined-up telemetry and configurations are essential to keep the twin trustworthy.
- Model realism. Enough radio network and traffic detail is needed to change decisions without burning unnecessary computational resources.
- Resource commitment: Significant time, money and resources will need to be committed to digital twin and AI projects, with the benefits taking time to materialise.
Furthermore, it will be crucial for operators to ensure robust internal governance processes, once a digital twin is implemented. This will include clear promotion paths from twin to production, explicit rollback and audit trails for AI-driven changes. The encouraging trend is that common patterns and AI-native simulation stacks are emerging, which in turn should lead to safe automation becoming a more routine path than a risky experiment.
Finally, getting the timing of implementing digital twins right will be critical for a good return on investment. Whilst many operators are nearing the end of the 5G investment cycle, having tried-and-tested models up and running before the next big investment will be essential and may be key to making new opportunities, such as the deployment of 6GHz spectrum or 6G technology, work better in practice.
Summary
Digital twins are not a silver bullet. But paired with AI, they are already starting to deliver measurable benefits. A winning pattern seems to be emerging: gather relevant and meaningful telemetry data, use the twin to learn by going through a cycle of “simulate, propose, deploy, sync” and then utilise AI to carry out meaningful analysis effectively and efficiently to provide a clear framework for decision makers to take action. By using such a robust framework, increased network autonomy becomes a series of bounded, testable and value-generating steps rather than a leap of faith.
Footnotes
[1] Fabricated data that behaves like real data, used to train and test models when real data is limited
[2] As an example, companies such as Aira Technologies are promoting the use of their GenAI-powered tools and resulting productivity benefits actively, see: https://www.aira-technology.com/rangpt-the-worlds-first-llm-based-utility-for-ran/
[3] For more detail, see: https://www.ericsson.com/en/press-releases/2/2025/2/chunghwa-telecom-and-ericsson-harness-ai-and-digital-twins-to-secure-network-performances-during-data-surge
[4] Telefónica reported on its original achievements in this post: https://www.telefonica.com/en/communication-room/blog/level-4-autonomy-digital-twin-o2-transport-network-germany/
[5] For more detail, see: https://inform.tmforum.org/features-and-opinion/how-verizon-is-using-digital-twins-to-reduce-energy-costs
