CASE STUDY
According to the GSMA, global mobile data traffic could increase almost fivefold by 2030. More devices, richer applications, cloud services, video and AI workloads will all place additional demands on radio access networks – the infrastructure connecting users and devices to the wider network.
For Latin America, that challenge intersects with another reality: operators need to expand and improve connectivity while controlling the cost and energy requirements of increasingly sophisticated infrastructure.
Countries such as Colombia are already examining what the next generation of mobile networks will require. A study by Colombia’ s Communications Regulation Commission( CRC) highlights spectral efficiency, energy efficiency, distributed processing and artificial intelligence as important elements in the evolution of mobile networks.
The broader lesson is that AI does not necessarily have to mean more computation. Used effectively, it can also mean less wasted computation.
Traditional RAN architectures were designed around relatively predictable workloads, with resources allocated according to established configurations. That model becomes harder to sustain as traffic patterns become more volatile.
That is the promise of AI-RAN: integrating artificial intelligence into the radio access network so that computing and network resources can be optimised in response to changing conditions.
GlobalLogic, the digital engineering company owned by the Hitachi Group, has been testing what happens when AI and GPU acceleration are brought into a 5G Open RAN environment. Its proof-of-concept trials produced striking results: processing times improved by as much as 15 times under certain conditions, while power consumption fell by around 40 %.
A second experiment produced an even more dramatic result. Using deep-learning models to distinguish genuine network-access attempts from those caused by noise or interference, GlobalLogic reports a 99.98 % reduction in false detections.
The figures come from laboratory tests rather than commercial deployments, so they should not be interpreted as guaranteed gains for operators. But they point towards a potentially significant shift in how mobile networks are designed and operated.
A stadium can suddenly generate a huge spike in demand. An industrial site can produce thousands of machine connections. A cloud application can shift workloads rapidly. An AI service can create new requirements for lowlatency connectivity and computing.
AI can help operators respond to those changes dynamically, potentially improving spectrum utilisation, energy efficiency and network performance.
“ Among its possibilities are a more efficient use of the spectrum, a reduction in energy consumption, better management of network resources and the ability to respond dynamically to changes in traffic,” says Sebastián Bainer, Senior Vice President and Head of Latin America, GlobalLogic.
For operators, the attraction is therefore not AI for its own sake, but the prospect of extracting more capacity and efficiency from increasingly complex infrastructure.
AI workloads are particularly suited to GPUs because they can perform large numbers of calculations in parallel. That makes accelerated computing attractive not only for AI applications but also for telecommunications workloads requiring substantial processing power. www. intelligentcio. com
INTELLIGENT CIO LATAM
15