It’s no secret that U.S. water infrastructure is aging and falling apart, and we’re losing the fight.
In fact, we’re hemorrhaging about 6 billion gallons of treated water every single day through broken pipes. That’s not a typo—6 billion gallons. Every. Single. Day.
While communities struggle to choose between paying their water bills or buying groceries, we’re literally watching our most precious resource drain into the ground.
But here’s what keeps me up at night. We’ve known about these problems for decades.
The American Society of Civil Engineers keeps handing out failing grades like they’re going out of style: a “C-” for drinking water, a “D+” for wastewater. We’re looking at a $1 trillion price tag just to fix what we’ve neglected. While politicians argue about budgets, 250,000 water main breaks happen every year.
Yet there’s a glimmer of hope in this mess, and it’s coming from an unexpected place: AI. That’s right. Last week, we talked about the drain and strain of water resources thanks to data centers that power artificial intelligence, but I don’t want to neglect the hope it offers in solving some of our biggest water issues as well.
What can I say? Life is full of contradictions.
The Hidden Struggle
Let’s review the nation’s drinking water infrastructure. It’s made up of more than 2 million miles of underground pipes and operated by nearly 150,000 public water systems.
About half of country’s water pipes are more than 45 years old. Some of these cast-iron pipes have been in the ground since the 1800s. On top of that, we’ve got 9.2 million lead service lines still carrying water into homes—pipes that are literally poisoning us. The infrastructure we’re depending on wasn’t built for today’s climate, today’s population, or today’s chemicals.
And it gets worse. Small leaks, the ones that don’t make the evening news, account for 30-50 percent of all water loss. They’re invisible killers of our water supply, draining resources while communities face water scarcity and/or water boil notices.
Traditional leak detection? It’s labor-intensive, skill-dependent, and frankly, outdated. We’ve been sending crews out with listening devices, hoping they can hear a leak through concrete, traffic noise, and a hundred other sounds. It’s like finding a needle in a haystack while wearing earplugs.
Enter AI-Powered Leak Detection
Now, here’s where things get interesting. Companies are using AI-powered acoustic sensors that can listen to water pipes 24/7 and identify leaks with stunning accuracy. I’m talking about systems like FIDO Tech’s tool that uses GPT-4 to analyze acoustic data in seconds, identifying the unique signature of a leak without needing to know anything about the pipe material, depth, or size.
Think about that for a second. A sensor can be placed on a water pipe, and AI does all the heavy lifting. It doesn’t get tired. It doesn’t need years of training to distinguish between the sound of a leak and traffic rumbling overhead. The technology has been deployed in London, Arizona, and Mexico, finding leaks that would have otherwise gone undetected for months or years.
And it’s not just about finding leaks; it’s about prioritizing them. These systems can rank leaks by size, so utilities know which ones to fix first. That means limited repair budgets get used where they’ll have the biggest impact. That’s smart resource management, something we desperately need.
But AI isn’t just reactive, it’s also predictive. Scientists at the University of Vermont and Utah State University developed something revolutionary, taking the federal government’s National Water Model and supercharged it with AI and real-time sensor data. Now it doesn’t just forecast stream flow, but it can predict water quality.
This matters because when a storm hits, utilities can know in advance how it’s going to affect their water supply. Will there be dangerous turbidity levels? Will algal blooms explode? Will sediment overwhelm treatment plants? Instead of scrambling after the fact, water managers get lead time to shut down vulnerable sources, issue warnings, and protect public health.
They tested this tool in New York City’s water supply system, a massive, complex network serving millions, and it worked. The AI successfully predicted when turbidity levels would spike, giving operators time to adjust operations before water quality became dangerous. This technology can be deployed nationwide, giving every community better tools to protect their drinking water.
“With the first ever application of the National Water Model to predict water quality, we’ve opened a new window that can really benefit the country as a whole moving forward,” said Andrew Schroth, the study’s lead researcher in a statement.
Seeing the Invisible
Here’s another breakthrough that should be making headlines. Researchers are using AI-powered vision systems combined with cameras to monitor water pollution in real-time. These systems can detect algal blooms, oil spills, industrial waste runoffs, dead fish, synthetic foams—you name it. The AI spots the pollution, identifies it, and can automatically alert authorities to take immediate action.
No more waiting for lab results. No more gaps in monitoring. Just continuous surveillance that catches problems when they start, not weeks later when the damage is done. For communities living near industrial sites or dealing with agricultural runoff, this could be life-changing.
And it goes deeper. Scientists are deploying AI systems with UV spectroscopy and fluorescence sensors that can detect chemical contamination in real-time. We’re talking about the ability to identify emerging contaminants, hundreds of thousands of chemicals in use today that aren’t part of traditional water quality tests. The AI establishes what “normal” looks like for each water system, and when something changes, it flags it immediately.
Filling Data Gaps
One of the biggest problems in water management is that we don’t have enough data, especially in vulnerable communities. Many rural areas and disadvantaged neighborhoods lack the monitoring infrastructure that wealthier communities take for granted.
AI is changing that equation. Machine learning models can predict water quality conditions in areas without sensors by analyzing data from nearby monitored sites, weather patterns, land use, and dozens of other factors. In Japan, researchers trained AI on data from 211 river catchments and achieved 91 percent accuracy in predicting river discharge in ungauged watersheds.
That means communities that couldn’t afford extensive monitoring networks can still get accurate water quality assessments. That means environmental justice, ensuring every community, regardless of zip code or income level, has access to information about their water safety.
AI systems are also helping identify lead service lines without physically inspecting every single pipe. Utilities can submit data to machine learning platforms that predict which service lines are most likely to contain lead, allowing them to prioritize replacements and meet their deadlines for getting lead out of the system. Every lead pipe removed is a child protected from neurological damage.
Cities like Toledo, Ohio, have received EPA grants to use these machine learning predictive models to assess lead service lines based on existing parcel and neighborhood-level data Utah State Today. The technology was pioneered in Flint, Michigan by researchers from the University of Michigan and Georgia Tech, where it was estimated to save $10 million in unnecessary excavations.
Smart Water Management
We’re not just talking about band-aids here. AI is fundamentally changing how we manage water resources. Platforms like Waterplan are creating dynamic, continuously updated water risk assessments that combine satellite imagery, local measurements, government reports, scientific papers, and real-time monitoring, all analyzed by AI that can understand context across different languages and data formats.
This context is crucial because water risks don’t stay put. Climate change is making droughts longer, floods more severe, and contamination events more frequent. We need systems that can adapt as fast as conditions change. AI models continuously learn, integrate new data, and adjust their predictions in real-time, exactly what we need for dynamic threats.
And here’s the kicker. AI doesn’t replace human expertise; it amplifies it. The best systems combine machine speed with human judgment, using expert validation to ensure accuracy while dramatically cutting the time from data collection to actionable recommendations from months down to hours.
What This Means for Your Community
If you’re thinking this all sounds great but wondering when it’ll reach your neck of the woods, here’s what you need to know. The technology already exists. It’s not science fiction. It had been deployed in cities across the U.S. and around the world.
But deployment requires investment and political will. It requires utilities to embrace new technology. It requires policymakers to fund not just pipes and treatment plants, but smart monitoring systems. It requires all of us to demand better.
The 2021 Bipartisan Infrastructure Law allocated $55 billion for water infrastructure, and recent EPA funding brought another $11.5 billion to states in 2024. That’s progress, but experts say we need $109 billion annually in the next 20 years to truly fix our water systems. We need to ensure those dollars go toward not just replacing old pipes, but building smart, resilient systems.
Look, I’ve spent my career fighting for communities whose water has been poisoned by corporate negligence and government failure. I’ve seen what happens when we ignore infrastructure, when we kick the can down the road, and decide certain communities aren’t worth protecting.
AI isn’t a silver bullet. Nothing is. We still need massive investment in physical infrastructure. We still need stronger enforcement of water regulations. We still need political leaders who will prioritize clean water above special interests.
But AI gives us tools we’ve never had before. It gives us the ability to detect problems early, predict crises before they happen, monitor continuously instead of sporadically, and make smart decisions about where to invest limited resources. It democratizes access to sophisticated water quality information, helping level the playing field for all communities.
Clean water isn’t a luxury. It’s a right. And in a country as wealthy and technologically advanced as ours, there is no excuse for the water crisis we’re facing.
We have the technology. We have the knowledge. What we need now is the will to act.
You can start the conversation in your community by asking your local water utility to explore AI-powered monitoring and leak detection. Contact your elected officials and insist they prioritize water infrastructure funding.
Test your own water—don’t wait for someone else to tell you it’s safe. And most importantly, don’t accept that this crisis is inevitable. We can make progress on these issues.
Great article! Lots of useful information esp about the use of AI to aid in the water systems. This needs to hit the news!
That which is here attributed to "AI" can be easily accomplished using multi decades old acoustic signal processing techniques coupled with decades old machine learning. Why does this matter? Because LLMs require gobs of power- data centers and all that. And you, Erin, know what data centers cost in terms of water and greenhouse gas emissions.
But it's good to know that tools are available.