Stargate LLC

Based on Wikipedia: Stargate LLC

Half a Trillion Dollars to Build the Future

In January 2025, President Donald Trump stood at a White House podium and announced what he called "the largest AI infrastructure project in history." The numbers were staggering: up to five hundred billion dollars to be invested in artificial intelligence infrastructure across the United States by 2029. The project's name, Stargate, evoked both science fiction grandeur and the ambition to open portals to technological worlds that don't yet exist.

The scale invited immediate comparison to the Manhattan Project—the secret World War II program that built the first atomic bombs. Whether Stargate will prove as transformative remains to be seen, but the comparison tells you something about how its backers want the world to perceive it.

The Players at the Table

Stargate isn't a government program. It's a joint venture—a business arrangement where multiple companies pool resources toward a shared goal—incorporated in Delaware as Stargate LLC. The founding partners represent a fascinating cross-section of the technology industry.

OpenAI, the company behind ChatGPT, brings its artificial intelligence expertise and will handle what Sam Altman, its CEO, calls the "operational responsibility." SoftBank, the Japanese conglomerate known for making enormous bets on technology companies, provides the financial muscle and contributes its CEO, Masayoshi Son, as chairman. Oracle, Larry Ellison's enterprise software giant, supplies cloud computing infrastructure. And MGX, an investment firm backed by the United Arab Emirates, rounds out the founding group.

The technology partners extend further: Arm, the British chip design company whose architecture powers nearly every smartphone on Earth; Nvidia, whose graphics processing units have become the essential hardware for training AI systems; and Microsoft, OpenAI's longtime backer and cloud computing partner.

This last relationship is particularly interesting. Microsoft and OpenAI signed a new agreement alongside the Stargate announcement that gives Microsoft "right of first refusal" on OpenAI's future cloud computing needs. In plain terms, this means OpenAI must offer Microsoft the opportunity to host its computing workloads before turning to competitors—but if Microsoft declines or can't meet the demand, OpenAI can look elsewhere. Stargate represents that "elsewhere."

What Are They Actually Building?

Data centers. Lots of them.

Modern artificial intelligence systems require enormous computing power, both to train the underlying models and to run them once trained. A single query to an AI chatbot might seem instantaneous to you, but behind the scenes it involves calculations across thousands of specialized processors running in vast, climate-controlled warehouses. These facilities consume electricity on an almost industrial scale.

Larry Ellison, speaking at the announcement, said ten data centers were already under construction in Abilene, Texas. The plans extend beyond Texas to other states and eventually to countries including the United Kingdom, Norway, Japan, and the United Arab Emirates.

By September 2025, OpenAI announced five additional sites: Shackelford County in Texas, Doña Ana County in New Mexico, Lordstown in Ohio, Milam County in Texas, and a midwestern location they haven't publicly disclosed. The company said these additions bring Stargate to nearly seven gigawatts of planned capacity. To put that number in perspective, seven gigawatts could power roughly five million average American homes. The full commitment is ten gigawatts by the end of 2025.

The Money Question

Five hundred billion dollars is an almost incomprehensible sum. It exceeds the annual GDP of countries like Norway or Austria. So where is it coming from?

The venture launched with an initial commitment of one hundred billion dollars. According to reporting by The Information, SoftBank and OpenAI each pledged nineteen billion dollars initially and would each hold forty percent ownership in the joint venture. Oracle and MGX would each contribute seven billion dollars. The remainder would come from limited partners—essentially outside investors—and debt financing, which means borrowed money.

Not everyone was convinced the funding was real. Elon Musk, whose company xAI competes directly with OpenAI, publicly claimed the venture lacked the financing to meet its promised investment levels. Sam Altman denied this. Arm's CEO Rene Haas called the financial backing "quite solid."

The Wall Street Journal offered a more nuanced picture: SoftBank, despite being the primary financial backer, might collect only ten percent in actual equity funding, with the rest coming from debt or debt-like arrangements. This distinction matters. Equity means ownership—money invested in exchange for a share of the company. Debt means loans that must be repaid with interest. A heavily debt-financed project carries different risks than one funded primarily through equity.

In April 2025, SoftBank announced its first ten billion dollars would be borrowed from Mizuho, one of Japan's largest banks, along with other lenders. By May, JPMorgan Chase agreed to lend 2.3 billion dollars to OpenAI and its partners for the Abilene data center projects. These facilities are being developed through a separate joint venture involving Blue Owl Capital, Crusoe Energy Systems, and Primary Digital Infrastructure, and will be leased to Oracle.

Then came August 2025. Bloomberg reported that the project had not actually started and no funds had been raised to meet the initial five hundred billion dollar budget. Market uncertainty, American trade policy, and fluctuating valuations for AI hardware had caused delays. The same report noted that Masayoshi Son had earlier approached TSMC, the Taiwanese semiconductor manufacturer, about investing one trillion dollars in Arizona—suggesting that Son's ambitions may sometimes outpace available capital.

Going Global

Stargate was announced as an American project, but its ambitions quickly expanded internationally.

In April 2025, the Financial Times reported that OpenAI and Oracle were considering future investments in the United Kingdom. By May, something more concrete emerged: "UAE Stargate," a data center planned for the United Arab Emirates with an expected opening in 2026. The partners for this project include Nvidia, Cisco, G42 (an Emirati AI firm), Oracle, SoftBank, and OpenAI.

The UAE connection deserves particular attention. MGX, one of Stargate's founding investors, is backed by UAE sovereign wealth. G42 is deeply connected to the Emirati government. The Gulf states have been aggressively positioning themselves as AI hubs, offering cheap energy, minimal regulation, and sovereign capital. Stargate UAE represents a significant expansion of American AI infrastructure into that ecosystem.

In October 2025, OpenAI announced Stargate Argentina—the first Stargate site in Latin America. Located in Patagonia, that cold and sparsely populated region at the southern tip of South America, the project involves a partnership with Sur Energy and an estimated investment of up to twenty-five billion dollars. The facility will have capacity of up to five hundred megawatts, making it the largest data center in Latin America, surpassing existing facilities in Brazil, Chile, and Mexico.

Why Patagonia? Cold climates reduce the enormous cooling costs that data centers incur. Argentina also offers relatively cheap electricity and a government eager for foreign investment. The location might seem remote, but for computing workloads that don't require millisecond response times—like training AI models—physical distance matters less than power costs and cooling efficiency.

The Hardware Race

Building data centers is only half the challenge. You also need to fill them with the right chips.

Nvidia has dominated the market for AI accelerators—the specialized processors that train and run neural networks. Their GPUs, originally designed for rendering video game graphics, turned out to be remarkably well-suited for the parallel mathematical operations that machine learning requires. This accident of history made Nvidia one of the most valuable companies in the world.

But Stargate isn't putting all its chips in one basket. AMD, Nvidia's longtime competitor in the graphics processor market, will supply up to six gigawatts worth of its Instinct GPUs. The agreement includes a remarkable provision: if AMD meets certain milestones, OpenAI may buy a ten percent stake in the company. This would give OpenAI direct ownership in its hardware supplier—an unusual vertical integration play that could reshape the AI supply chain.

Broadcom, another chip company, will supply ten gigawatts of custom hardware. Unlike AMD and Nvidia, which sell general-purpose processors, Broadcom specializes in application-specific integrated circuits—chips designed for particular tasks rather than general computing. Custom AI chips can be more efficient for specific workloads, though they lack the flexibility of general-purpose processors.

The Promise and the Skepticism

At the January announcement, Larry Ellison made a bold claim: Stargate could lead to AI-facilitated production of mRNA vaccines against cancer. Such vaccines, he said, could be designed "robotically"—by which he meant leveraging AI—"in about 48 hours."

This claim merits some unpacking. mRNA vaccines work by instructing cells to produce proteins that trigger an immune response. The COVID-19 vaccines from Pfizer and Moderna used this technology. Cancer vaccines, still largely experimental, would work similarly—training the immune system to recognize and attack cancer cells. The design of such vaccines requires analyzing a patient's specific tumor mutations, which is computationally intensive and theoretically could benefit from AI acceleration.

Whether forty-eight-hour design cycles are realistic depends on many factors beyond computing power: regulatory approval, clinical testing, manufacturing capacity. Ellison's timeline may be aspirational rather than practical.

Trump announced he would use emergency declarations to expedite Stargate's development, particularly regarding energy infrastructure. This signals the enormous power demands these facilities require and the regulatory obstacles that typically slow large infrastructure projects. Whether emergency powers are appropriate for what is, at its core, a private commercial venture raises questions that extend beyond technology into governance and the relationship between government and industry.

The venture claims it will create more than one hundred thousand American jobs. This number is difficult to verify and likely includes both direct employment and indirect economic effects. Data centers themselves employ relatively few people once built—they're largely automated facilities. The job creation claims probably factor in construction workers, support services, and the broader economic activity that flows from such large investments.

What Stargate Isn't

The name "Stargate" actually predates this venture. In 2024, OpenAI and Microsoft used the term for an earlier project: a one hundred billion dollar AI supercomputer. That initiative, also named after the 1994 science fiction film where stargates were portals to other worlds, was apparently folded into or superseded by the larger joint venture announced in 2025.

The film reference is fitting. In both the movie and subsequent television series, stargates represented shortcuts—ways to traverse vast distances instantaneously. The Stargate LLC project similarly promises shortcuts: accelerated vaccine development, faster AI progress, economic transformation. Whether those promises materialize or remain science fiction depends on whether the funding actually flows, whether the technology delivers, and whether the geopolitical environment remains stable enough for such massive long-term investments to proceed.

The Bigger Picture

Stargate represents something larger than any single infrastructure project: the race to achieve artificial general intelligence, or AGI—AI systems that match or exceed human cognitive capabilities across all domains. OpenAI has stated this as its explicit goal. The company believes that whoever develops AGI first will hold unprecedented technological and economic power.

The United States, China, and to a lesser extent the European Union and Middle Eastern states are all competing in this race. Stargate is, in part, an attempt to ensure American dominance by building the computing infrastructure that advanced AI development requires. The emergency declarations, the sovereign wealth fund participation, the international expansion—all reflect the strategic importance that governments and investors assign to AI leadership.

Whether five hundred billion dollars over four years is enough, whether the venture's complex ownership structure will prove manageable, whether the technical advances will materialize—these remain open questions. But the ambition is unmistakable. Stargate aims to build the physical foundation for whatever artificial intelligence becomes in the coming decades. Whether it succeeds or joins the long list of technology projects that promised more than they delivered, its announcement marks a moment when the scale of AI investment shifted dramatically upward.

The portals are being built. Where they lead remains to be seen.