Edge computing: proximity as a new competitive advantage

In this scenario, telecoms operators possess an asset that US hyperscalers cannot easily replicate: the extensive reach of their networks and their proximity to users. Thousands of telephone exchanges, transmission towers and points of presence spread across the country represent potential edge nodes that are already connected and powered. In September 2024, Deutsche Telekom signed an agreement with Google Cloud for the co-deployment of 1,500 multi-access edge computing nodes, establishing a partnership model in which telecoms providers supply physical proximity and network assets, whilst hyperscalers contribute software and distribution channels.

Italia is also taking steps in this area. In June 2026, the Department for Digital Transformation within the Prime Minister’s Office signed an agreement with thirteen Italian universities to launch trials on edge-cloud computing. The aim is to assess how distributed processing can improve the quality and performance of the public sector’s digital services, from healthcare systems to urban transport services.

The universities involved – including the Politecnico di Milano, Sapienza, Bologna and Napoli Federico II – will be working on practical use cases, ranging from telemedicine to smart traffic management.

The competitive advantage of the edge is not limited to latency. Processing data locally reduces transmission costs to the cloud, improves resilience in the event of connectivity outages and enables sensitive information to be kept within controlled perimeters – a crucial aspect for regulated sectors such as healthcare, finance and defence. According to Scale Computing, edge deployments reduce operating costs by 60–80 per cent compared with traditional architectures, thanks to automation and operational simplification.

For artificial intelligence, the edge represents the frontier of distributed inference. Whilst model training still requires centralised data centres with thousands of GPUs, the execution of already trained models can take place on much more compact hardware, close to the end user. It is the difference between teaching a neural network to recognise a face (training) and using that capability to unlock a smartphone (inference). The latter operation can run on a dedicated chip the size of a postage stamp.

The challenge for Europe is not to remain a mere spectator of this transformation. American hyperscalers dominate the centralised cloud, but the edge opens up opportunities for local players capable of capitalising on their proximity, knowledge of the local area and relationships with institutions. Telecoms firms, national system integrators and utilities with distributed infrastructure have assets at their disposal that the Big Tech giants do not possess. The game has only just begun.