Meta's $10 Billion Bet in the Texas Desert

In October, it was a $1.5 billion project. Five months later, it's $10 billion. Same plot of land. Same desert. Completely different ambition.

Meta announced Thursday that it's increasing its planned investment in an AI data center in El Paso, Texas, by more than sixfold. The facility, targeting 1 gigawatt of capacity by 2028, will sprawl across 1.2 million square feet and require up to 4,000 construction workers at peak. It's Meta's third data center in Texas and, by any measure, its most audacious infrastructure play yet.

But the announcement landed on a brutal day for the company. Meta's stock dropped 7% on Thursday after two court losses related to child safety failures on Facebook and Instagram. The stock is now down 16% for the year. And earlier this week, the company confirmed hundreds of layoffs across recruiting, sales, global operations, and its VR division.

So why is Meta spending like it's flush when Wall Street is punishing it like it's broken?

The Structural Problem Nobody Talks About

Here's the uncomfortable truth about Meta's AI spending spree: unlike every major rival, it has no cloud business to offset the cost.

Google has Cloud. Amazon has AWS. Microsoft has Azure. Each of them can build a data center, flip it on, and charge enterprises for every compute hour. The infrastructure pays for itself — and then some.

Meta doesn't have that. Its entire $135 billion capital expenditure budget for this year — a figure it disclosed to Wall Street in January — must be funded by advertising revenue. Every dollar spent on servers in El Paso is a dollar that can't go to shareholders, buybacks, or dividends. That's why Wall Street is applying "extra scrutiny," as analysts politely put it.

The question isn't whether Meta can afford this. It probably can — the company generates enormous cash flow. The question is whether the AI infrastructure bet will ever produce a return that justifies the scale. And on that, the company has offered vision but not yet a clear business model.

Sponsored

Advertise with Us

[email protected]

Sponsored

What $10 Billion Actually Buys

The El Paso facility is designed to be a serious piece of infrastructure, not a press release. A 1 gigawatt data center is roughly equivalent to the power consumption of a mid-sized American city. For context, most hyperscale data centers operate in the range of 100–500 megawatts.

To fill it, Meta has been on a chip-buying spree. In February, the company signed massive supply deals with Nvidia and AMD. This week, it committed to becoming the first customer for Arm's new data center processor. And it just unveiled four new versions of its in-house MTIA AI accelerators — chips it first disclosed publicly in 2023.

The new facility will be liquid-cooled using a closed-loop system, which Meta says will keep water consumption comparable to "a typical golf course in the region." The company is also funding eight water restoration projects in Texas and partnering with nonprofit DigDeep to bring running water to more than 100 homes in the area for the first time.

That last detail isn't just good PR. It's a direct response to Meta's own history: in 2018, the company broke ground on a $750 million data center in Georgia, and reports later emerged that taps ran dry in the surrounding county. The backlash from communities near AI data centers has grown louder across the country, and Meta is clearly trying to get ahead of it.

The Bigger Race

This isn't just about Meta. The company's accelerated spending reflects an industry-wide conviction — shared by Google, Amazon, Microsoft, and a growing list of challengers — that whoever controls the most compute wins the AI era.

The strategic logic is straightforward, if unproven: AI models require massive compute to train and run. The companies with the most infrastructure can build better models, attract more users, and lock in enterprise customers. Falling behind now means paying a higher price to catch up later.

What makes Meta's position unusual is its motivation. It doesn't sell AI compute to others. It uses AI to make its advertising products smarter — better targeting, better recommendations, better content ranking. Every billion spent on data centers is, in theory, an investment in making Facebook and Instagram more effective ad platforms.

Whether 1 gigawatt of compute translates into meaningfully better ads — and whether better ads justify the cost — is the central question Meta's investors are sitting with right now.