A Smarter Way to Fix America’s Crumbling Water Pipes

Utilities test predictive AI to catch pipe failures early, cut costs, and rethink how aging water systems are managed

6 Feb 2026

A growing number of US water utilities are turning to predictive artificial intelligence to identify weak points in ageing pipe networks before failures occur, as decades-old infrastructure faces rising pressure and limited funding.

Eastern Municipal Water District, which supplies water to communities east of Los Angeles, has begun piloting predictive AI to forecast where water main breaks are most likely. The utility has partnered with VODA.ai to analyse network data and rank pipe segments by risk, allowing maintenance to be planned rather than triggered by emergencies.

The approach marks a shift for an industry traditionally focused on responding to bursts and leaks after they happen. During the pilot, the model assessed existing records, including pipe age, materials and historical break data, and produced a risk profile across the network.

Utility officials involved in the project say that when subsequent break events were reviewed, many occurred in locations the system had already identified as high risk. Roughly three quarters of recorded breaks aligned with the highest-risk segments flagged by the model. Several failures also occurred in pipes with no previous break history, exposing gaps in conventional planning methods that rely heavily on past incidents.

Managers say the ability to anticipate problems could reshape how maintenance programmes are run. Planned repairs are typically cheaper and faster than emergency work, and cause less disruption for customers. Utilities also point to reduced water loss, fewer road closures and lower complaint volumes as potential benefits, particularly at a time when budgets are under strain.

Pressure on water systems is increasing nationwide. Much of the US water network was built more than 50 years ago, and investment in replacement has lagged behind need. Predictive analytics is being promoted by technology providers and industry analysts as a way to help utilities prioritise scarce capital towards the highest-risk assets.

Adoption, however, is uneven. Predictive models rely on accurate and consistent data, which not all utilities possess. Smaller systems may also struggle with the cost and skills required to deploy new tools. As a result, early use has been concentrated among larger operators with more developed records.

Even so, projects such as EMWD’s are being closely watched. As data quality improves and pilots demonstrate measurable savings, predictive AI is expected to play a larger role in how utilities manage ageing infrastructure and plan long-term investment.