Inside Stargate: The $500 Billion Gamble That Could Define America AI Future

May 25 2025 09:59
Standing on 1,200 acres of red Texas dirt, watching 600-ton cranes move like mechanical dinosaurs against the vast sky, it's hard to imagine this barren landscape will soon house what could be the most powerful AI infrastructure on Earth. But that's exactly what's happening in Abilene, Texas, where a partnership between OpenAI, SoftBank, and Oracle is betting $500 billion on a project that sounds more like science fiction than business reality.


Project Stargate wasn't supposed to become public knowledge so quickly. The name started as an internal code word at OpenAI, partly because early data center designs resembled the iconic portal from the science fiction series. But when President Trump announced the project the day after his inauguration, calling it "the most important project of this era," Stargate transformed from tech industry speculation into a national priority.

The numbers are staggering. Eight planned buildings will house up to 400,000 of Nvidia's latest Blackwell chips, creating one of the world's largest known computing clusters. The power requirements alone dwarf most infrastructure projects: 1.2 gigawatts, enough electricity to power 750,000 homes. To put that in perspective, asking ChatGPT a question already uses about 10 times more energy than a Google search. Stargate will amplify that by orders of magnitude.

Chase Lochmiller, the founder of Crusoe Energy Systems and the man overseeing construction, describes the project with the enthusiasm of someone who climbed five of the world's seven tallest mountains. "What we're undergoing right now is sort of the largest infrastructure build in human history," he tells me, gesturing toward the 155 vehicles currently moving dirt around the clock. "This is like when we built the interstate highway system, but for intelligence."


The selection of Abilene, a town of about 125,000 people known more for cattle ranching than cutting-edge technology, reveals the practical realities driving AI infrastructure development. While Silicon Valley and Seattle have the talent, they lack something Stargate desperately needs: massive amounts of cheap, available electricity.


West Texas has become a renewable energy powerhouse, with wind farms stretching across the horizon. But all that clean electricity generation created an unusual problem: more supply than demand. "We came to Abilene honestly because it's a community that's very rich in low-cost and clean energy," Lochmiller explains. "What they lack is actually demand for power."

This energy abundance comes with local economic benefits that help explain why Abilene agreed to an 85% property tax reduction for Stargate. Mayor Weldon Hurt, who makes a dollar a year in his role and runs a pest control business on the side, frames it simply: "We're getting 15% of billions of dollars. That's still millions of dollars on an annual basis."

But the tax incentive reveals a broader pattern across America, where smaller communities are offering aggressive deals to attract data center projects, hoping to participate in a tech boom that largely bypassed them over the past two decades.


The partnership behind Stargate brings together three very different entities with aligned interests. OpenAI needs massive computing power to train increasingly sophisticated AI models. SoftBank's Masayoshi Son wants to make what he calls "the most important investment of this era." Oracle sees an opportunity to compete with Amazon and Microsoft in cloud infrastructure by specializing in AI-optimized data centers.

Sam Altman, OpenAI's CEO, describes the evolution of their computing needs with remarkable candor:

The urgency becomes clear when Altman describes recent challenges with their image generation feature, which became so popular that he joked on social media about GPUs "melting." While the metal wasn't literally melting, the demand surge forced OpenAI to borrow computing capacity from research projects and slow down other features. "If we had more GPUs, we would be more able to handle demand surges like this," Altman explains. "More compute means we can give you more AI."


Perhaps the most uncomfortable aspect of Stargate is its environmental impact. Every major tech company has pledged to achieve net-zero carbon emissions by 2030, but the AI boom is pushing energy consumption in the opposite direction. Individual server racks at Stargate will consume 130 kilowatts each, compared to 2-4 kilowatts for traditional data centers just 20 years ago. Lochmiller is refreshingly honest about this contradiction:

The optimistic view, which Lochmiller shares, is that AI will eventually accelerate the development of new energy technologies like small modular reactors and nuclear fusion. The massive energy demand from AI could provide early adoption opportunities for these emerging technologies. But that's a bet on future innovation solving problems created by current expansion.

Crusoe has implemented some innovative solutions, including a closed-loop cooling system that uses the same million gallons of water indefinitely, rather than consuming millions of gallons daily like traditional data centers. But the fundamental energy equation remains challenging.


Walking through downtown Abilene, the impact of Stargate feels both immediate and abstract. Construction has brought 2,200 workers from seven different states to the site, creating a visible economic boom. Hotels are booked, restaurants are busier, and there's a palpable sense of excitement about being part of something historic.

But conversations with locals reveal mixed feelings. A computer science student at Abilene Christian University tells me she loves seeing jobs come to Abilene but worries about AI's broader implications: "The big picture to me is AI's gonna take over and then where are we as humans? It honestly just terrifies me."

Others are more optimistic. A local resident mentions that the project could provide alternatives to oil and gas jobs, potentially helping address substance abuse issues in the region. "This adds something different and hopefully breaks some of those cycles," he says.

The job creation promises remain somewhat vague. Mayor Hurt mentions numbers ranging from 400 to 1,200 permanent positions, but admits uncertainty: "I don't know a lot about AI centers or how many people they employ. I'm gonna be honest about that." Data centers, once operational, typically require far fewer workers than during construction.


Not everyone believes Stargate will reach its ambitious goals. The $500 billion price tag has drawn particular skepticism, with observers noting SoftBank's history of overpromising on major investments. Even Son acknowledges past mistakes: "Sometimes I get crazy over excited and I make mistake. Like WeWork and you know, I have many scars."

Recent developments have added to the uncertainty. DeepSeek, a Chinese AI company, recently demonstrated that sophisticated AI models can be built with significantly less computing power than previously thought. This raised questions about whether massive infrastructure projects like Stargate are actually necessary.

It's a bet that demand will always expand to fill available supply. Altman's response is telling:

The financial sustainability question looms large. OpenAI ran a $5 billion loss in 2024, and there's no clear path to profitability that justifies the scale of investment. Microsoft has already begun pausing some data center projects, suggesting the industry may be approaching limits.


Stargate isn't just about business; it's about national competitiveness. The project represents America's response to China's growing AI capabilities, particularly after DeepSeek demonstrated that innovative approaches can compete with massive resource investments.

Former diplomat Anja Manuel frames the stakes clearly: "We're in a real race. The Chinese are giving us a run for their money. China's right behind us. We're very, very close. Fifty percent of the world's AI researchers are Chinese."

President Trump's tariff policies add another layer of complexity. Much of the material needed for data center construction comes from China, potentially increasing costs. The AI supply chain remains deeply international, with sophisticated chips manufactured primarily in Taiwan and Korea.

This creates both challenges and opportunities. Higher costs could slow expansion, but they might also accelerate efforts to build more domestic manufacturing capacity. The CHIPS Act has already begun bringing some advanced semiconductor production to the United States.


Standing in what will soon be an AI data hall, surrounded by the infrastructure that will house 50,000 chips, it's impossible not to wonder what we're really building. Altman admits the uncertainty: "Do I think I could have sat here in 1905 and told you what we were about to discover in physics and that 40 years later we would have an atomic bomb? Definitely not."

The comparison to nuclear physics isn't accidental. Like the Manhattan Project, Stargate represents a massive bet on technology whose ultimate implications we can't fully understand. The optimistic vision includes AI systems that accelerate scientific discovery, solve climate change, and create new forms of human flourishing. The pessimistic version involves job displacement, energy consumption spiraling out of control, and power concentrated in the hands of a few technology companies.

What seems certain is that decisions made in places like Abilene over the next few years will shape the trajectory of AI for decades to come. The physical infrastructure being built today will determine what kinds of AI systems are possible tomorrow. As Lochmiller puts it, watching another 600-ton crane swing into position against the Texas sky:

Time will tell if he's right, but one thing is certain: we're about to find out what happens when the world's most ambitious technological bet meets the reality of physics, economics, and human nature.