The exponential growth of artificial intelligence (AI) is transforming consumer & B2B uses, as well as requirements on telecoms infrastructures. Telecom operators are now at a crucial crossroads, where the evolution of their fixed and mobile networks is essential to support this technological revolution.
AI is also a formidable tool for optimizing network management and performance, but this issue falls within the scope of the “AI for Networks” topic.
Network evolution challenges
The main challenges facing network infrastructures for AI applications are increasing bandwidth, reducing latency, and handling very large volumes of connected objects, whether fixed or in motion. There is a strong association between network configuration and the efficiency of AI systems.
Many use cases reinforce the attractiveness of 5G networks, which is why it has even been said that AI is the “killer application” of 5G!
Data sovereignty constraints will also have to be considered and will have an impact on the localization of data processed by AI.
- Increased Bandwidth: to deliver sufficient bandwidth, investment in fiber and 5G networks is essential.
Let's not forget that training Large Language Models (LLM) of AI requires the transfer of huge quantities of data: several Tera bytes for text databases.
The development of multimodal AI, able to manage images, video and sound, will also see data traffic rise exponentially.
The use cases are numerous: from virtual or augmented reality applications to Smart Cities and their networks of continuously streaming video cameras.
In the US, operators Verizon and AT&T have already begun deploying 5G networks with 10Gbps data rates to support AI applications.
In its annual Cisco Visual Networking Index report, Cisco announces that by 2025, mobile data consumption is expected to reach 77 exabytes per month.
And by 2030, mobile data traffic is expected to quadruple.
- Low latencies: These requirements are set to rise and will be critical for certain AI-enabled services. This is true in the consumer market, with cloud gaming applications, for example, or in the B2B market, in smart factory production lines or drone-based solutions. To achieve these low latencies, 5G's flow slicing and prioritization capabilities can be used. And to further reduce latency, operators can bring AI processing infrastructures closer to users, by deploying Edge Computing solutions. Amazon Web Services, which has implemented Edge solutions, has been able to reduce latency by up to 50% in certain applications.
- Massive IoT: the number of connected objects in an area requiring AI will also increase massively, with smart cities and factories, the development of connected eyewear for the mass market, and increasingly autonomous vehicles. 5G will make it possible to handle this high density of connected objects.
- Data sovereignty: in addition to reducing latency, the deployment of domestic Edge Computing infrastructures, on which data will be processed by AI, can also ensure the sovereignty of sensitive data. Most hyperscalers and other AI companies now offer such solutions. Beware, however, of the investments required, and business plan assumptions will need to be carefully reviewed.
SOFRECOM’s expertise
In such a challenging environment, Sofrecom can play a decisive role. As a consulting and engineering firm, we offer end-to-end expertise in designing, deploying and maintaining flexible, scalable and open network infrastructures and services.
Sofrecom's expertise supports telcos in tackling the technical and strategic issues related to AI systems deployment.
Thanks to tailor-made solutions, operators can not only meet current requirements, but also anticipate future market developments.
Conclusion
Telecom operators face major challenges to adapt their networks to be able to cope successfully with the rise of artificial intelligence. By working with trusted partners like Sofrecom, they can turn these challenges into opportunities!
