The AI Race Is Becoming a Power Race

An 11-gigawatt private power complex isn’t a normal data center expansion.

It’s the kind of electricity output you’d expect from multiple nuclear reactors.

Yet one is being built in West Texas right now to guarantee enough electricity for one of the largest AI campuses ever planned.

A few weeks ago in the Daily Disruptor, we said that electricity, not chips, is becoming the biggest bottleneck in the AI boom.

Projects like this are a response to that issue.

Instead of waiting for utilities to expand the grid, the biggest tech companies in the world are now starting to secure their own power at industrial scale.

In Amarillo, Texas, the plan is for a massive data center campus that comes with its own power generation. The early build is anchored by natural gas turbines, but its long-term roadmap also includes nuclear energy.

Ultimately, the plan is to build enough power on site so the project isn’t stuck waiting years for transmission upgrades and grid interconnection approvals.

Because AI companies are no longer assuming that the power grid will keep up.

So they’re making sure they don’t have to depend on it.

More Power, Please

The project in Amarillo is tied to Fermi America and what’s being called Project Matador.

This isn’t a single warehouse filled with servers. It’s an 18-million-square-foot campus built around its own power backbone, with long-term capacity that could reach roughly 11 gigawatts.

For context, that’s the kind of output you associate with several large power plants operating together. It’s more electricity than many states add in an entire planning cycle.

And it’s already taking shape.

Construction timelines have already been set, stretching into 2026 and beyond. Developers have secured major permitting approvals in Texas, including a large Clean Air permit tied to gas generation on site. And the cooling systems are being engineered as hybrid dry-wet towers to manage water use at scale.

Image: Fermi America

The reason this facility is being built today is simple. Data center demand is growing faster than the U.S. power grid is being upgraded.

Across the country, power companies are warning that it takes years to plug new projects into the grid. In some areas, building new transmission lines alone can take five to seven years.

Bringing a closed nuclear plant back online can take even longer. It has to be inspected, repaired, updated and approved before it can produce power again.

And when many companies try to buy gas turbines at the same time, there simply aren’t enough to go around. Orders can take years to deliver.

But the AI race isn’t slowing down to allow for permitting cycles and manufacturing delays.

As we recently noted in the Daily Disruptor, hyperscalers have turned the AI race into a spending race. Estimates for combined spending by the largest players in 2026 run into the hundreds of billions of dollars. That money is flowing into chips, land, buildings and network infrastructure.

It’s also flowing into power. And not just in Amarillo.

Microsoft has signed long-term nuclear power agreements tied to U.S. generation. Amazon has secured dedicated capacity near existing nuclear facilities. Google is backing advanced reactor partnerships aimed at bringing new generation online later this decade.

What’s different about the Texas project is its scale.

Instead of negotiating for incremental supply, it’s being designed as a self-contained energy system attached to AI compute.

That changes the economics.

Image: resources.org

Electricity is one of the biggest costs for a large data center. Because training and running AI models uses a lot of power.

When energy prices rise, profits shrink.

But if a company can build or lock in its own power supply, it can protect itself from price swings. It also makes it easier to plan future growth when you know that electricity will be there when you need it.

And if you can control when power comes online, then you control when compute comes online.

That matters in a competitive race where moving faster gives you an edge.

And it matters even more now that Washington has stepped in.

In his State of the Union address last month, President Trump said big tech companies should build or pay for their own power plants to support new AI data centers. He called it a “ratepayer protection pledge.” According to Trump, his goal is to make sure regular households aren’t stuck with higher electric bills because of AI demand.

Since then, the White House invited major tech firms to meet about the plan, including companies like Amazon, Google, Microsoft, Meta, Oracle and OpenAI.

Whatever your view of the politics here, this is a real issue that needs to be addressed.

Communities are realizing that a single AI campus can demand as much electricity as a mid-sized city. That creates pressure on local grids, and it raises obvious questions about who pays for upgrades.

Projects like Amarillo offer one compelling solution.

Build it yourself.

Here’s My Take

We’re starting to see a change in how the largest technology companies in the world think about infrastructure.

For years, the grid was treated as a shared platform that would scale alongside demand.

But AI is testing that assumption.

As growth in compute starts outpacing growth in power generation, the companies with the capital and urgency to solve this problem can’t wait for the system to adjust.

They’re being forced to build around it.

If you’re a member of Strategic Fortunes, none of this should come as a surprise.

Last month, I recommended a company positioned to profit from the coming surge in electricity demand tied to AI.

And as projects like the one in Amarillo start taking shape, that thesis is playing out exactly as expected.

If you’re not a member of my flagship service yet…

What are you waiting for?

Click here to find out more about Strategic Fortunes.

Regards,

Ian KingChief Strategist, Banyan Hill Publishing

Editor’s Note: We’d love to hear from you!

If you want to share your thoughts or suggestions about the Daily Disruptor, or if there are any specific topics you’d like us to cover, just send an email to [email protected].

Don’t worry, we won’t reveal your full name in the event we publish a response. So feel free to comment away!