The network questions I am asked to settle

When a supplier comes to be treated as high-risk, the technical work is less about the new equipment than about sequence. I separate the decision that the authorities and the board own, which is whether and how fast to remove the supplier, from the engineering decisions that follow: which functions come out first, what runs in parallel and for how long, and what else changes on each site visit. At Corvalta we moved the packet core first, running old and new side by side for four months, and then replaced radios on 4,100 sites in 23 clusters, each with a performance baseline and a rollback plan. I insist on three things suppliers dislike: logs and configuration that we own and can read without them, exit support priced at signature, and a declared interest from anyone in the scoring who has worked for a bidder, including me. Supply-chain security is mostly patience and paperwork, and it fails where either runs out.

Electricity is one of the largest costs a mobile operator can still change through engineering, and the radio network uses most of it: about three-quarters of Corvalta's 362 gigawatt hours in 2025. I treat energy per gigabyte as an engineering measure with an owner in each region, not as a figure for the sustainability report. The levers are well known: switching off legacy layers, putting carriers and symbols to sleep at night, removing cabinets that need air conditioning, refarming spectrum to more efficient technologies and buying power under long-term contracts. Our ten-year agreement with Solenna Rinnovabili, a solar developer, now covers about a fifth of our consumption. The harder part is deciding what to give up. Sleep modes cost some peak capacity at night, the early 3G switch-off cost us customers and a power purchase agreement fixes a price for a decade. I put each of those costs in front of the executive committee beside the saving.

I have negotiated a radio-sharing agreement from scratch in Portugal and inherited one in Italy, and in both the engineering was the easy part. The difficult questions are about governance: who decides upgrade timing on a shared site, how a partner can veto work in its busiest weeks, how costs are split when one party wants 5G and the other does not yet, and how the arrangement ends. At Corvalta our rural agreement with Sestriva Telecom covers 2,600 sites, half hosted by each operator. Spectrum is the other scarce input. A challenger rarely has the low-band holdings of older operators, so I spend time on refarming, on how fixed-wireless homes consume mid-band capacity in the evening and on the technical evidence we give the commercial team before licence decisions. I describe spectrum positions in general terms in public; holdings, renewals and their prices belong in the board's papers.

I distinguish resilience that has been designed from resilience that has been demonstrated. After our July 2024 voice outage I stopped reporting the number of redundant systems and started reporting live tests passed and failed. We switch live traffic between our two core sites every quarter at night, check nightly for configuration differences between them and give the results to the audit and risk committee, including the February 2025 test that interrupted calls for 160,000 customers. At the mast I classify sites by what they serve, such as fixed-wireless homes, the only coverage for a village or a hospital, and give those classes longer battery autonomy than the general standard. Heatwaves are now a design condition in southern Italy, both for cooling at core sites and for longer power cuts at radio sites, and I plan for them as a recurring event rather than an exception.

For a mobile challenger that resold copper broadband, the switch-off of copper is an engineering question as much as a commercial one. Each migrating customer can go to wholesale fibre, with an installation visit at a third party's pace and a lower margin, or to our own 5G fixed-wireless service, which is cheaper to provide but consumes the mid-band capacity our mobile customers use in the evening. I set a cap on fixed-wireless homes per radio sector from our busy-hour model, require an indoor signal test before a customer is offered the service and review the cap twice a year, sector by sector. I also treat power as part of the product. A fibre or fixed-wireless line stops in a power cut unless it has a battery unit, so we state the minutes of autonomy plainly and equip customers who depend on telecare or alarm devices.