Making cities inclusive, safe, resilient and sustainable is Goal 11 of the UN 2030 Agenda. Yet in Italy more than 42% of the water fed into distribution networks is still lost, and in 2024 abstractions for drinking use hit their lowest level in twenty-five years. Between the national recovery plan (PNRR), the Internet of Things and predictive algorithms, how will Italian municipalities guarantee drinking water in their residents’ homes?
With the PNRR, Italy has chosen to safeguard water resources so as to guarantee drinking water in citizens’ homes, also through the use of AI systems for an effective, efficient and sustainable management of water services. How will Italian municipalities make such important choices?
The challenges facing Italian municipalities
The climate crisis and global warming pose an unprecedented challenge: making major administrative choices to secure drinking water and fight water stress. Historically the emergency was managed through short-sighted supply choices — from limited delivery by tankers to rationing; today AI systems allow different, strategic choices and, above all, tackle the central problem: the losses that occur in the aqueducts, thanks also to Internet-of-Things (IoT) technology.
The most recent figures show the scale of the problem. According to ISTAT, total distribution losses stand, on the national average, at 42.4% of the volume fed into the network (2022 figure), peaking at 45.5% for in-house operators. The ISTAT focus released in March 2026 notes that “in 2024, 8.87 billion cubic metres of water were abstracted in Italy for drinking use: the lowest level in the last 25 years”; that in 2024 more than one million residents in the provincial capitals (5.8% of the population) were affected by water-supply rationing; and that “in 2025, 2.7 million households reported irregularities in the supply service”.
The PNRR and the digitalisation of networks
Onto this front came the PNRR investment devoted to “reducing losses in water-distribution networks, including the digitalisation and monitoring of networks” (M2C2, investment 4.2). As communicated by the Ministry of Infrastructure and Transport, “a total of €1,900 million has been allocated to 103 interventions”, with the stated goal “of building at least 45,000 km of water network at district level by 31 March 2026”. District metering and the digital monitoring of networks are, indeed, the technical precondition for predictive algorithms to work: without sensors and data, no AI can detect a leak.
A virtuous example: the “Acquedotto 4.0” in Santarcangelo di Romagna
New AI technologies can certainly help support choices aimed at achieving the 2030 Agenda goals. A virtuous example is the Municipality of Santarcangelo di Romagna, which in 2021 launched the “Acquedotto 4.0” project together with the University of Bologna, the Hera Group and Rezatec. The results of the trial, initially run on 500 km of network, were disclosed by the Hera Group: “the algorithm made it possible to identify the 35% of the Santarcangelo network on which 69% of the breakages occurred”. In other words: by concentrating maintenance on little more than a third of the network, the vast majority of breakages are prevented. On the strength of these outcomes the project was extended to further municipalities, covering some 2,800 km of network.
Nor is this an isolated case among operators: predictive maintenance of water networks has become one of the most mature application areas of AI in local public services, precisely because the benefit is measurable and the stakes — water — are a common good by definition.
Water as a testing ground for the smart city
A systemic note is in order. Water management is perhaps the most concrete area of application of the so-called smart city: here AI does not raise the dilemmas of public-space surveillance, but it still demands data governance, information quality and long-term infrastructure investment. Technology alone does not repair the pipes: far-sighted administrative choices and a public stewardship able to measure outcomes are needed.
Conclusions
The Italian water paradox lies entirely in the numbers: the lowest abstraction level in twenty-five years coexists with network losses above 42% and more than one million citizens subject to rationing. The path is set, between PNRR-funded district metering and predictive maintenance entrusted to algorithms; the first results, as the experience of Santarcangelo di Romagna shows, are there. In the light of the above, one wonders whether local authorities and operators will manage to turn experimentation into ordinary practice, carrying the PNRR investments beyond the pilot phase and translating them into stable district-metered networks, sensors and predictive maintenance, so as to guarantee citizens — starting with the most fragile territories — the most essential of services: drinking water in their own homes.
