The AI boom is forcing a new sustainability conversation - one measured in megawatts, gallons, materials, and operating resilience.

Artificial intelligence may be digital.

Its footprint is anything but.

Every AI model sits on top of a growing physical system: data centers, substations, transmission lines, cooling equipment, backup generation, fiber networks, semiconductor fabs, and increasingly enormous amounts of electrical infrastructure.

That Creates An Uncomfortable Reality for the Intelligence Economy:

The faster AI grows, the more important the physical efficiency of the infrastructure beneath it becomes.

For years, sustainability in technology was often discussed through carbon commitments, renewable-energy purchases, and corporate ESG reports.

Those things still matter.

But AI is changing the conversation.

The Next Generation of Sustainability Will Increasingly Be About Something Much More Operational:

How much intelligence can we produce from the infrastructure resources we have?

And that could make sustainability one of the most underestimated competitive advantages in AI infrastructure.

The Sustainability Equation Is Changing

The Old Sustainability Conversation Sounded Something Like This:

How do we reduce the environmental footprint of technology?

The Emerging AI Infrastructure Conversation Is Different:

How do we continue scaling compute when power, water, land, equipment, and grid capacity are increasingly constrained?

That's a very different problem.

It means efficiency isn't simply an environmental objective anymore.

It's an infrastructure strategy.

A data center that can deliver more compute from the same electrical capacity has an advantage.

A cooling system that reduces water dependence can open markets where water availability is politically or environmentally sensitive.

A campus capable of integrating renewables, storage, nuclear, natural gas, fuel cells, or other generation technologies has more options.

A region capable of supporting infrastructure growth without overwhelming its grid, water systems, or communities becomes more attractive.

This is why sustainability belongs among the eight pillars of the Infrastructure of Intelligence™.

It interacts with almost every other one.

The AI Power Problem Is Also an Efficiency Problem

Everyone is talking about how much electricity AI will require.

And they should be.

Data center electricity consumption is expected to grow significantly as AI workloads expand.

But there are two sides to that equation.

Side One: Build More Power.

Utilities add generation.

Transmission gets expanded.

New substations are constructed.

Developers explore nuclear, natural gas, renewables, batteries, fuel cells, and behind-the-meter generation.

Side Two: Get More From Every Megawatt.

That side may prove just as important.

AI Infrastructure Is Pushing Innovation Across:

• Advanced cooling

• Power management

• Higher-voltage architectures

• Server efficiency

• GPU utilization

• Energy storage

• Heat reuse

• Microgrids

• Grid-interactive data centers

• More efficient chips and computing architectures

Put Simply:

The cheapest megawatt may eventually be the one you don't need to build.

When power availability is becoming one of the biggest constraints on data center development, efficiency suddenly becomes very valuable.

Cooling Is Entering the Spotlight

For decades, air cooling handled much of the data center industry's thermal management.

AI changed the math.

High-density GPU racks can generate enormous amounts of heat, pushing operators toward more advanced cooling technologies.

That's why liquid cooling has become one of the most important infrastructure technologies to watch.

But there's an important distinction.

Liquid cooling does not automatically mean high water consumption.

Direct-to-chip systems, closed-loop cooling architectures, immersion technologies, dry coolers, and hybrid systems can dramatically change how facilities manage heat.

The sustainability opportunity isn't simply using less water.

It's designing cooling systems around local conditions.

A cooling strategy that makes sense in Virginia may not make sense in Arizona.

A system optimized for abundant water may be inappropriate in a drought-prone region.

That means the future of AI cooling will likely become increasingly regional.

And that brings us to something I think the industry needs to talk about more.

Sustainability Is Becoming a Site-Selection Issue

Imagine two communities competing for the same billion-dollar AI infrastructure project.

Both have land.

Both have fiber.

Both have economic-development incentives.

But one can demonstrate:

Reliable power.

Water resilience.

Advanced wastewater infrastructure.

Renewable generation.

Energy storage.

Fast permitting for distributed generation.

Heat-reuse opportunities.

Community support.

And a long-term resource plan capable of supporting additional campuses.

Which location looks less risky?

Exactly.

This is why sustainability is moving beyond the corporate ESG department.

It is becoming part of infrastructure underwriting.

Developers need to know whether resources will remain available.

Utilities need to know whether demand growth can be supported.

Investors need to understand long-term operating risk.

Communities need to understand the impact of large infrastructure projects.

And hyperscalers need confidence that campuses built today will still make sense twenty years from now.

The Twin Infrastructure Principle

Here's another piece of the sustainability conversation I believe deserves more attention.

AI infrastructure rarely arrives alone.

A Major Data Center Campus Can Trigger Investment In:

• Transmission.

• Generation.

• Substations.

• Fiber.

• Roads.

• Water infrastructure.

• Wastewater systems.

• Housing.

• Workforce development.

• Industrial suppliers.

That creates what we call the Twin Infrastructure Principle.

When private AI infrastructure development is coordinated with public infrastructure investment, both systems can become stronger.

A transmission upgrade built partly because of industrial demand can strengthen regional electrical capacity.

A wastewater project designed around new industrial users can modernize community infrastructure.

Fiber built to serve a data center corridor can improve regional connectivity.

New generation can expand local energy resources.

Sustainability, therefore, isn't necessarily about stopping infrastructure development.

Sometimes it's about designing infrastructure development so that more than one party benefits from it.

That's the opportunity.

The Next Big Sustainability Market: Waste

There is another resource AI infrastructure produces in enormous quantities.

Heat.

Lots of it.

Data centers take electricity, run computing equipment, and ultimately convert much of that energy into heat that must be removed.

Historically, that heat has largely been treated as waste.

That may begin changing.

Across colder climates especially, developers and municipalities have explored ways to redirect data center heat toward:

District heating.

Greenhouses.

Industrial processes.

Aquaculture.

Buildings.

Agricultural applications.

Not every location will make the economics work.

But the broader idea matters.

The Intelligence Economy may eventually become much better at turning infrastructure waste streams into infrastructure inputs.

Heat is one example.

Water reuse is another.

Battery systems can interact with grids.

Backup generation can potentially evolve into grid-supporting assets.

Even data center campuses themselves could become participants in regional energy systems rather than simply enormous electricity consumers.

That's a much more interesting sustainability story.

Follow the Innovation

Whenever infrastructure encounters a constraint, capital starts looking for a solution.

That's exactly what is happening here.

The sustainability challenge surrounding AI is creating markets around:

Advanced nuclear.

Small modular reactors.

Fuel cells.

Long-duration energy storage.

Grid software.

Liquid cooling.

Water recycling.

Heat reuse.

High-efficiency electrical equipment.

Microgrids.

Renewable generation.

Carbon-free firm power.

And increasingly efficient computing hardware.

This is why I wouldn't view AI sustainability only through an environmental lens.

View it as an emerging infrastructure technology market.

Some of the companies solving these problems could become important second-order winners of the Intelligence Economy.

Companies to Watch

Schneider Electric

Power management, cooling, electrical infrastructure, and data center efficiency increasingly converge as facilities become denser.

Eaton

Electrical equipment may not sound like a sustainability technology.

Until you realize that managing, distributing, protecting, and optimizing electricity is central to improving infrastructure efficiency.

Vertiv

As rack densities rise, thermal management becomes mission-critical. Cooling efficiency is becoming directly connected to how much compute operators can deploy inside a facility.

NVIDIA

The sustainability conversation starts upstream.

More computing performance per unit of energy can have enormous implications when multiplied across millions of accelerators.

Microsoft, Google, Amazon and Meta

The hyperscalers are increasingly becoming infrastructure laboratories.

Power procurement, nuclear agreements, cooling technologies, renewable development, carbon removal, water strategies, and energy storage initiatives coming from these companies can influence how the broader industry evolves.

Watch what they deploy - not just what they pledge.

The Hot Take

AI's sustainability problem may ultimately become one of its greatest innovation catalysts.

Scarcity forces optimization.

Power constraints encourage efficiency.

Water concerns accelerate cooling innovation.

Grid congestion encourages distributed generation.

Carbon targets create demand for new energy technologies.

Community concerns force developers to think more carefully about infrastructure integration.

And rising capital costs reward designs capable of producing more compute from fewer physical resources.

The industry isn't going to stop building AI infrastructure because resources are constrained.

It's going to spend enormous amounts of money figuring out how to build around those constraints.

That is where the opportunity is.

What to Start Asking

When evaluating an AI infrastructure project, company, or market, don't stop at:

"How sustainable is it?"

Start Asking:

• How many megawatts does it require per unit of compute?

• Where does the power come from?

• How much water is actually consumed rather than simply circulated?

• What cooling architecture is being used?

• Can wastewater or reclaimed water be utilized?

• Could waste heat become economically useful?

• Can the facility participate in grid services?

• What happens when the campus doubles in size?

• Can the surrounding community's infrastructure scale with it?

Those questions reveal much more than an ESG score ever could.

The Golden Nugget

Here's the sustainability metric I would start watching closely:

Compute per Megawatt.

The AI infrastructure race is usually framed around who can secure the most electricity.

But over time, another competition could become equally important:

Who can produce the most intelligence from every available megawatt?

If power remains constrained, improving that equation could unlock enormous economic value.

And the winners won't only be data center operators.

They'll include the chipmakers, cooling companies, electrical equipment manufacturers, utilities, energy developers, engineering firms, and infrastructure technologies helping make it possible.

The IOI Take

The sustainability pillar isn't separate from the AI infrastructure story.

It connects the entire system.

Power determines what can operate.

Water determines how certain systems can cool.

Land determines where infrastructure can be built.

Policy determines what development is permitted.

Capital determines what solutions are economically viable.

Sustainability determines whether the entire system can continue scaling.

The regions and companies that understand this early won't simply build greener AI infrastructure.

They may build better AI infrastructure.

And in an Intelligence Economy increasingly constrained by physical resources, that distinction could become extremely valuable.

Infrastructure of Intelligence | Sustainability Pillar

The Infrastructure of Intelligence™ tracks the eight physical systems shaping the Intelligence Economy: Power, Connectivity, Land, Water, Workforce, Policy, Capital, and Sustainability.

Artificial intelligence may be digital.

Its future is physical.

Workforce

Policy

Capital

Sustainability