Project Stargate — What Is the $500B OpenAI Mega Data Center in Texas?

A $500 billion bet on AI is taking shape in the red dirt of West Texas. It’s called Project Stargate — and if you haven’t heard the name yet, you likely will. OpenAI, SoftBank, and Oracle are building what could become one of the world’s largest AI compute clusters in Abilene, Texas. Here’s what’s actually happening, why it’s there, and what it means for the future of AI.

What Is Project Stargate?

Project Stargate is a multi-site AI infrastructure initiative led by OpenAI, SoftBank, and Oracle. The first — and most talked-about — site is in Abilene, Texas, where construction is already under way. The project is being built out by Crusoe, a data center startup whose founder, Chase Locke Miller, is overseeing the Abilene campus.

So far, the partners have committed $100 billion to the effort and have said they are prepared to invest up to five times that as they expand. The scale is what grabs headlines: we’re talking hundreds of thousands of GPUs, multiple buildings, and power needs measured in gigawatts.

Why Abilene, Texas?

Abilene isn’t a random pick. West Texas offers cheap, abundant power — much of it from wind — and communities eager for large-scale investment. Data centers like Stargate need huge amounts of electricity; a single rack of the latest AI chips can draw dozens of times more power than a traditional server rack. Abilene had the land, the grid potential, and the willingness to negotiate. In exchange, the city agreed to incentives that include a large cut in property taxes on the project, betting that the remaining tax revenue and economic spillover will still be a net win.

The first phase at Abilene is sometimes referred to as “Project Ludicrous” — a nod to the breakneck schedule. The goal is to have the first site operational by mid-2026, with construction running around the clock and workers drawn from multiple states.

What’s Actually Being Built?

On the ground, Stargate in Abilene means eight large buildings on a 1,200-acre campus, each designed to house massive clusters of GPUs — the chips that train and run today’s AI models. We’re not talking ordinary servers; these are NVIDIA Blackwell–class GPUs, built for AI training and inference. When fully built out, the site could hold on the order of 400,000 chips, making it one of the largest known AI compute clusters in the world.

Power is the other headline number. The site is being wired for about 1.2 gigawatts of capacity — enough to run hundreds of thousands of homes. Cooling is another challenge: advanced GPUs run hot, so the facility uses a closed-loop water system that Crusoe says is filled once and then reused, rather than consuming millions of gallons per day like some older data center designs.

Why So Much Money and So Much Build?

Demand for AI compute has exploded. ChatGPT and other products showed that people want to use AI far more than many insiders initially planned for. Training bigger models and serving more users means more chips, more power, and more buildings. Sam Altman and others have been clear: they see this as the kind of infrastructure moment that defines an era — similar in spirit to building out the backbone of the internet or the interstate system, but for “intelligence” as a utility.

SoftBank’s Masayoshi Son has framed it as a bet on artificial general intelligence (AGI) and the idea that the first movers in compute will shape what comes next. Not everyone agrees that the right answer is “build as much as possible.” DeepSeek and other efforts have raised the question of whether similar AI capability can be achieved with less compute. Some big tech firms have also paused or scaled back data center plans. So Stargate exists in a live debate: Is this the right scale, or are we overbuilding?

Jobs, the Town, and the Skeptics

For Abilene, Stargate means construction jobs now and a smaller number of long-term roles once the data centers are running — security, maintenance, cooling, and operations. Data centers are not like old-fashioned factories; they don’t employ tens of thousands of people on the floor. So the lasting job count is uncertain, and locals have mixed feelings: hope for growth and new opportunities, and concern about power use, water, and how much the town is really giving up in tax breaks.

There’s also broader anxiety about AI replacing jobs elsewhere. OpenAI’s leadership has said they try to be upfront: AI will change many jobs, remove some, and create others, and the transition will feel uneven. Stargate is a symbol of that bet — that the demand for AI will justify the cost and the build.

The Bottom Line

Project Stargate is real, it’s big, and it’s under construction. Whether the full $500 billion ever gets deployed across all planned sites is an open question; what’s clear is that OpenAI, SoftBank, and Oracle are putting serious capital and engineering into making the Abilene site a cornerstone of their AI infrastructure. How smart AI can get, how much compute it “needs,” and whether this scale pays off will play out over the next several years.

If you want to see the site, hear from the people building it, and get more context on the partnerships and the math, the video below goes deep on Stargate, Abilene, and the broader AI infrastructure race.

Watch the Full Report

For a detailed look at the Stargate build, the partners involved, and the view from the ground in Abilene, watch the full segment here:

Project Stargate — Inside the $500B OpenAI Mega Data Center in Abilene (Bloomberg)