top of page

AI May Live in the Cloud. But Its Infrastructure Doesn't.

  • 5 days ago
  • 3 min read

Artificial intelligence can feel almost invisible. We type a question, click a button, and an answer appears seconds later. But somewhere behind that simple interaction is a very physical piece of infrastructure: a data center filled with servers, cooling equipment and electrical systems. And as AI grows, so does the infrastructure behind it.


That raises an interesting question:

Are we building the water and energy infrastructure needed to support it?




The Water We Don't See

Most discussions about data centers and water focus on cooling. That's understandable. Servers generate heat, and many data centers use water as part of their cooling systems.


But there's another water connection that's much easier to overlook: Electricity.


A new analysis released by Ceres looked at data centers across seven major U.S. data-center states: Virginia, Texas, California, Illinois, Georgia, Ohio and Arizona. Together, those states host about half of all U.S. data centers.


According to the report, 78% of the electricity generated in those states came from power plants that use water to operate.

And by 2030, Ceres estimates that as much as 72% of the total water consumed as a result of U.S. data-center operations could come indirectly from generating their electricity. 

In other words, the water footprint of a data center doesn't necessarily stop at its property line.


This Isn't One City's Question

Across the country, communities, utilities and data-center developers are beginning to wrestle with the same question:

How much additional infrastructure can existing water and electricity systems support?


That's especially important because many data centers are being built in clusters.

One facility may be manageable. Ten, twenty or fifty facilities concentrated in the same region become a very different infrastructure question. And geography matters.

A gallon of water used in a water-rich region doesn't necessarily have the same impact as a gallon used in an area already experiencing water stress.


Ceres found that 66% of the electricity generated using water in the states it studied came from areas facing medium-high to extremely high water stress. 


That's one reason we're seeing questions about data-center growth in very different places across the United States—from Nebraska and New Mexico to Texas, Arizona and Virginia. Recent reporting shows communities examining everything from water availability to electricity demand as projects grow larger and more numerous. This isn't about whether AI is good or bad. AI is here, and the infrastructure supporting it will continue to grow.


The more useful question is:

How do we build that infrastructure intelligently?


Virginia Offers an Interesting Example

Loudoun County, Virginia, is home to one of the country's largest concentrations of data centers. Its water utility began building a reclaimed-water distribution system in 2010 to provide non-potable water for irrigation and industrial cooling—including the county's rapidly growing data-center industry.

Today, that system includes approximately 20 miles of reclaimed-water pipelines.

In 2025 alone, Loudoun Water delivered more than 750 million gallons of reclaimed water to customers, helping save an equivalent amount of potable drinking water.

That's the part of this story we find particularly interesting. Data centers need cooling.

But cooling doesn't necessarily require drinking-quality water. If highly treated wastewater can safely perform that job, why use potable water when another water source is available?


Water and Energy Are Becoming One Conversation

A few weeks ago, we wrote about European rivers falling so low that power generation itself was threatened. This story turns that relationship around. Water affects our ability to produce energy. And producing energy can affect our water supply.


As electricity demand grows—from data centers, manufacturing, transportation, buildings and everything else we increasingly electrify—those two systems become harder to think about separately. That doesn't mean we stop building. It means the infrastructure supporting growth has to evolve along with it:

  • Reclaimed water.

  • More efficient cooling.

  • Closed-loop systems.

  • Better water storage.

  • Different energy sources.

  • Smarter decisions about where facilities are built.


There isn't one solution. But there is a common principle:

Don't automatically use drinking-quality water when the job doesn't require drinking-quality water.


That's not really an AI lesson. It's a water lesson. And it's becoming increasingly important as we build the infrastructure of the future.


💧 Water Thought of the Week

Innovation doesn't eliminate our need for water. It makes managing it intelligently even more important.


At Expiatech, we believe the future of water isn't finding more—it's making better use of the water we already have.



Sources
 
 
bottom of page