Two very different fights are playing out in Texas.

Near Amarillo, developers are pitching what could become one of the largest artificial intelligence campuses in the country. Eighteen million square feet. Talk of multiple nuclear reactors. Power generation on a scale more typical of a major city.

In San Antonio, residents are pushing back against large data facilities being built near homes and schools. Their concerns are more immediate. Diesel backup generators. Noise. Air quality. Cumulative industrial impacts in growing neighborhoods.

The projects are different. The tension underneath them is the same.

How do we build the digital infrastructure this country needs without repeating the environmental mistakes of the past?

Let’s start with something that is too often left unsaid. We need data centers.

Artificial intelligence systems, cloud computing, financial markets, medical research, defense operations and basic internet services all rely on them. Compute power is becoming strategic infrastructure, as essential to economic competitiveness and national security as ports, rail lines and power plants. If we do not build it here, it will be built elsewhere.

The question is not whether to build. The question is how.

The Amarillo proposal tells us something about where the industry is headed. Scale. The mention of four large nuclear reactors suggests ambitions measured in gigawatts. That is industrial magnitude. It reflects the explosion in AI demand and the race to secure reliable power.

But nuclear is not the inevitable solution.

It comes with enormous capital costs, long timelines and substantial water use. It centralizes risk and locks communities into infrastructure that will shape local economies for generations. And it diverts attention from cleaner technologies that can be deployed faster and more flexibly.

At the same time, the San Antonio debate is a reminder that siting matters. When industrial-scale facilities land next to residential neighborhoods, people notice. Backup diesel generators may only run periodically, but testing and outages still produce emissions. Cooling systems create constant background noise. Communities deserve more than after-the-fact explanations. They deserve transparent environmental review and meaningful input before permits are granted.

Sustainability cannot be reduced to accounting.

Buying renewable energy credits does not eliminate the strain a massive facility places on the local grid. Carbon capture promises do not solve today’s emissions. Calling something “clean” does not make it so.

If a data center consumes one gigawatt of electricity, that electricity must come from somewhere. The solution should be new renewable generation, not offsets.

Every viable square foot of rooftop on a large data campus should be covered in solar panels. Rooftop solar alone will not power a multi-gigawatt AI complex. But at this scale it can generate significant daytime electricity and reduce peak demand. Leaving industrial rooftops empty while marketing climate leadership sends the wrong message.

Beyond rooftops, developers should pair large solar and wind projects with long-duration storage. Wind and solar complement each other in places like the Texas Panhandle. Storage technologies are improving rapidly. Diesel backup systems may be familiar, but familiarity is not a sustainability strategy.

Water is another piece of the puzzle. Cooling large data centers can require millions of gallons a day. In water-stressed regions, that is a serious concern. Dry or hybrid cooling systems should be standard wherever feasible. Reclaimed wastewater should be prioritized over potable supplies. Desalination is not a free pass. It is energy intensive and can harm marine ecosystems. Diverting water from agriculture or residential users is not sustainable planning.

If a region cannot supply industrial water demand without harming existing communities, that region may not be the right site for a water-intensive facility.

None of this is anti-growth. It is strategic growth.

A responsible national strategy would encourage data centers to co-locate with new renewable projects rather than rely on paper offsets. It would push for long-duration storage instead of diesel. It would require transparent reporting of energy and water use. It would protect neighborhoods from cumulative industrial impacts.

The United States can expand artificial intelligence infrastructure and cut emissions at the same time. The engineering solutions exist. The economics are moving in the right direction. What is missing is a clear expectation that prosperity and responsibility go together.

The cloud may feel intangible. Its impacts are not.

These buildings consume real power and real water. They occupy real land next to real communities. The choices made today will last decades.

We can build the data centers we need. But we should insist on building them in a way that strengthens both our economy and the places people call home.