Data Centres and Water

Data Centres and Water: The Hidden Constraint Behind AI Infrastructure

After more than 20 years delivering infrastructure at scale, I can tell you this:

When systems start pushing limits, You begin to see dependencies that were always there.

But never mattered, until now.

Right now, AI infrastructure is starting to expose one of those dependencies.

And it’s not compute.

It’s not even power.

It’s water.

What People Think

As in my previous articles, most discussions around AI infrastructure focus on:

·         Compute

·         Power

·         Data centres

And increasingly, cooling.

But cooling is usually treated as a technical problem.

Something that can be solved with better systems.

Better design.

Better efficiency.

What gets missed is this:

Cooling isn’t just engineering.

It’s resource dependent.

A Quick Clarification (Why Water Matters)

Earlier, we talked about compute as processing power at scale.

Thousands of processors running continuously.

All of that processing generates heat.

And that heat has to be removed.

At smaller scales, air cooling can manage it.

At AI scale.

It can’t.

What’s Actually Happening

Modern data centres are moving toward:

·         Liquid cooling

·         Direct-to-chip cooling

·         Immersion systems

All of which rely heavily on water.

Because water is far more efficient at removing heat than air.

That’s not new.

What’s new is the scale.

Where It Breaks

On paper, cooling is solvable:

“Add more cooling capacity.”

In reality, this introduces a new constraint:

1. Water Availability

Not all locations have:

·         Reliable water supply

·         Sufficient volume

·         Sustainable access

And many of the locations with strong power access

Are already water constrained.

2. Environmental and Regulatory Pressure

Water usage is becoming a sensitive issue.

Particularly in:

·         Drought-prone regions

·         Urban growth areas

·         Environmentally regulated zones

This adds:

·         Approval delays

·         Operational restrictions

·         Public scrutiny

3. Competing Demand

Water isn’t just used by data centres.

It’s shared with:

·         Agriculture

·         Communities

·         Industry

Which means:

AI infrastructure is now competing for a critical resource.

Operator Insight: This Is How Constraints Stack

There are always patterns that determine infrastructure build success, the analysis for Datacentre builds would have a similar pattern, we see this pattern in infrastructure repeatedly.

Example:

You solve the primary constraint.

Power.

Then you hit the next

Cooling.

Then you realise:

Cooling depends on something else.

Water.

And suddenly, what looked like a solved problem.

Isn’t solved at all.

The Constraint Chain (This Is the Real Issue)

AI infrastructure is no longer dealing with single constraints.

It’s dealing with dependency chains.

These dependency chains are: Compute → Power → Heat → Cooling → Water

And every step adds complexity.

Every step introduces risk.

Where This Connects to the 5 Constraints

Water doesn’t sit outside your framework.

It sits inside it.

·         Power - drives heat

·         Land - determines water access

·         Capital - increases with advanced cooling systems

·         Workforce - specialised cooling expertise

·         Fibre - enables distribution to reduce load concentration

This is what real infrastructure looks like.

Interconnected.

What This Means for Industry

We’re already starting to see shifts:

·         Data centre location decisions influenced by water availability

·         Increased investment in water-efficient cooling systems

·         Growing scrutiny on environmental impact

 And importantly:

Water is becoming a design constraint, not just an operational input.

What Happens Next

Over the next few years, expect:

·         More innovation in low-water or waterless cooling

·         Greater regulatory pressure on data centre developments

·         Location strategies balancing power + water availability

·         Increased public attention on AI’s environmental footprint

And potentially:

Water becoming a limiting factor in certain regions.

Even where power is available.

Final Thought

In infrastructure, the constraint you ignore.

Is usually the one that slows you down.

Right now, everyone is focused on power.

Some are starting to look at cooling.

Very few are thinking about water.

But at scale.

That’s where the next pressure point is.

References & Further Reading

Footnote
This article is part of a series exploring the physical infrastructure behind the AI economy.

Disclaimer
The views expressed in this article are my own and are intended for general information and discussion purposes only. They do not represent the views of any employer, organisation, or client.

© 2026 Rodney Terry – Digital Backbone. All rights reserved.