Do these 3 things before closing this tab:
1Scan for outdated or missing drivers - takes under a minute2Repair Windows errors before they cause bigger problems3Fix the driver behind crashes, sound loss and screen glitchesProject Stargate is a private-sector, OpenAI-centered effort to build large AI data centers and supporting power infrastructure in the United States. It is not a $500 billion federal spending bill, not one Texas data center, and not proof that $500 billion has already been spent.
Announced on January 21, 2025, Stargate described a plan to invest $500 billion over four years, with $100 billion to be deployed immediately. The initial equity funders were SoftBank, OpenAI, Oracle, and MGX. The headline figure is best understood as an announced investment target, not a fully audited pool of cash.
As an Amazon Associate I earn from qualifying purchases.
What is Project Stargate?
Stargate is an infrastructure platform designed to give OpenAI access to substantially more computing capacity for training and running artificial-intelligence models. That capacity requires far more than servers: it includes land, buildings, NVIDIA accelerator systems, high-speed networking, storage, cooling, substations, transmission, generation, batteries, fiber, construction labor, financing, and long-term operations.
The original announcement described a new Stargate company. Later announcements used “Stargate” more broadly for a portfolio of data-center partnerships, leases, cloud arrangements, energy projects, and development agreements involving multiple companies. That broader usage is why the initiative’s reported capacity and investment figures vary by date and definition.
OpenAI is the principal compute customer and operational partner. SoftBank was identified as the financial lead, while Masayoshi Son was named chairman. Oracle, NVIDIA, Microsoft, CoreWeave, SB Energy, data-center developers, construction firms, infrastructure investors, and energy partners have participated in different parts of the buildout.
OpenAI’s original announcement said the project would initially focus on the United States, beginning in Texas, with a target associated with 10 gigawatts of capacity by 2029.
What the $500 billion means—and what it does not mean
| Figure | What it represents | Important qualification |
|---|---|---|
| $500 billion | Announced four-year investment target | Not evidence that this amount has already been spent or deposited in one project account |
| $100 billion | Amount described as being deployed immediately in the original announcement | Does not mean $100 billion had already been spent on January 21, 2025 |
| 10 gigawatts | Original U.S. capacity milestone | Gigawatts can refer to different stages, including planned, leased, powered, or operational capacity |
| More than $400 billion | OpenAI-reported investment associated with announced sites in September 2025 | May include planned commitments, leases, partnerships, and related development rather than audited expenditure |
| More than $450 billion | OpenAI-reported portfolio figure after the Michigan announcement in October 2025 | Company-reported and not a public independent accounting of cash invested |
The original announcement did not publish a complete financing schedule, capital-call timetable, campus-by-campus cost ledger, or independent audit of the full $500 billion. The figure may encompass different combinations of equity, debt, equipment purchases, leases, power assets, construction spending, and infrastructure commitments.
For personal-finance readers, the key distinction is between announced investment and deployed capital. Headlines often turn a long-term target into a statement that the money has already been invested. That is not established by the public announcements.
Who is involved?
| Company | Reported role |
|---|---|
| OpenAI | Principal compute customer and operational partner; uses the infrastructure for model training and inference. |
| SoftBank | Original financial lead; also involved in site, energy, and data-center development through related businesses. |
| Oracle | Major data-center and cloud-infrastructure partner, including operation of the Abilene infrastructure through Oracle Cloud Infrastructure. |
| MGX | Abu Dhabi-backed technology investment firm listed as an initial equity funder. A public dollar-by-dollar contribution breakdown was not provided. |
| NVIDIA | Supplies accelerated-computing hardware, including GB200-class systems and related networking technology. |
| Microsoft | Continues providing Azure cloud services to OpenAI while Stargate adds capacity through additional infrastructure. |
| CoreWeave | Specialized GPU-cloud and data-center partner involved in OpenAI-related capacity. |
| SB Energy | Develops powered infrastructure, generation, and storage for some Stargate-related campuses. |
No single participant supplies everything required. OpenAI brings demand for compute; investors bring capital; cloud providers supply operating expertise; NVIDIA supplies accelerators; developers and construction companies build the campuses; and utilities and energy partners provide the electricity infrastructure.
How far along is Stargate?
- January 2025: OpenAI announced the $500 billion, four-year plan, the immediate $100 billion deployment, the initial funders, and Texas as the starting location.
- July 2025: OpenAI and Oracle announced 4.5 gigawatts of additional U.S. data-center capacity. Combined with Abilene, OpenAI said the total would exceed 5 gigawatts and involve more than 2 million chips.
- September 2025: OpenAI announced five additional sites in Texas, New Mexico, Ohio, and the Midwest. It said the announced portfolio represented nearly 7 gigawatts of planned capacity and more than $400 billion of investment over three years.
- October 2025: The Midwest site was identified as Wisconsin, and OpenAI announced a more-than-one-gigawatt campus in Saline Township, Michigan. OpenAI said its broader portfolio exceeded 8 gigawatts and $450 billion of reported investment over three years.
- January 2026: OpenAI and SoftBank each invested $500 million in SB Energy. OpenAI also signed a 1.2-gigawatt lease for the Milam County, Texas, project.
- April 2026: OpenAI said Stargate had surpassed its initial 10-gigawatt U.S. milestone and that more than 3 gigawatts had been added during the preceding 90 days. It also said the Abilene site was operating with Oracle Cloud Infrastructure and NVIDIA GB200 systems.
- June 2026: Oracle, OpenAI, Related Digital, Blackstone, Walbridge, and Michigan officials marked construction of the Saline Township campus, known as “The Barn.” The announcement said more than 700 Michigan tradespeople were working there and that construction would create more than 2,500 union jobs.
The April 2026 capacity claim came from OpenAI and should be read as a company-reported milestone. It does not by itself establish that every gigawatt was fully built, grid-connected, powered, or running customer workloads.
Where are the major sites?
| Location | Reported status and participants |
|---|---|
| Abilene, Texas | Flagship site involving OpenAI, Oracle, Crusoe, and related infrastructure partners. OpenAI said NVIDIA GB200 racks arrived in 2025 and that early training and inference workloads began. Partial operation is not the same as full-campus completion. |
| Shackelford County, Texas | Part of the Oracle-related 4.5-gigawatt expansion. Capacity figures are primarily company-reported. |
| Milam County, Texas | OpenAI signed a 1.2-gigawatt lease, with SB Energy developing related powered infrastructure. A lease is not the same as live computing capacity. |
| Doña Ana County, New Mexico | Oracle’s “Project Jupiter,” expected by Oracle to include transmission lines, battery storage, and a dedicated substation. Employment figures are projections. |
| Lordstown, Ohio | SoftBank and OpenAI-related site described as being on track to become operational in 2026. “On track” is a forecast rather than proof of completion. |
| Wisconsin | Port Washington was announced with Oracle and Vantage. OpenAI also listed Mount Pleasant in a later community update involving Microsoft and OpenAI-related infrastructure. |
| Saline Township, Michigan | More-than-one-gigawatt campus known as “The Barn,” involving Oracle, Related Digital, Blackstone, OpenAI, and Walbridge. A construction milestone was marked in June 2026. |
These sites do not necessarily share the same ownership model. Some arrangements involve leases, some involve cloud operation, and others involve developers, financiers, utilities, or infrastructure partners. It would be inaccurate to say that OpenAI owns every site.
Crashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minutePC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Why power is the central constraint
A large AI campus is effectively an industrial electricity customer. Its requirements can include high-voltage grid connections, substations, transmission upgrades, onsite generation, battery storage, backup systems, and reliable power quality for dense computing equipment.
Oracle said its New Mexico project would include onsite transmission lines, battery storage, and a dedicated substation, with Oracle paying campus energy costs and related infrastructure under the project’s stated arrangement. OpenAI said the Michigan project would use existing excess transmission capacity and that project-funded upgrades would not be charged to local ratepayers. It described similar arrangements in Wisconsin and said SB Energy planned to fund and build generation and storage for the Milam County project.
Those are project-specific commitments, not proof that every Stargate site has solved its power requirements or that local electricity customers can never face indirect costs. The relevant questions are whether a utility has formally approved the arrangement, who pays if demand forecasts change, whether generation is physically dedicated or contractually purchased, and what happens to infrastructure if a campus is delayed or canceled.
Also, a gigawatt is a measure of power capacity, not a fixed number of GPUs. The usable compute depends on accelerator generation, rack design, utilization, cooling, networking, redundancy, and whether the figure describes facility load, IT load, or available computing capacity.
The Tool Desk
Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Water, cooling, land, and local communities
AI data centers produce heat and require cooling. OpenAI said the Michigan campus would use a closed-loop cooling system intended to significantly reduce water consumption. Closed-loop does not necessarily mean zero water use: facilities may still need water for initial filling, maintenance, humidity control, fire protection, and supporting systems. Exact consumption requires permits or utility records, not just a project description.
Construction also affects land use, transmission corridors, traffic, noise, air quality, backup generation, battery safety, and local tax arrangements. Project announcements emphasize jobs, training, grants, union construction, and protection of existing ratepayers. Those claims should be distinguished from independent measurements of economic or environmental impact.
Reported employment figures are not interchangeable. They include temporary construction jobs, permanent operations roles, indirect supply-chain jobs, and broader projected economic effects. For example, the Oracle expansion was associated with a company estimate of more than 100,000 jobs, the September 2025 site announcement cited more than 25,000 onsite jobs for a subset of sites, and the Michigan project announcement cited more than 2,500 union construction jobs. None of those figures should be presented as the number of permanent jobs created nationwide.
What role is the federal government playing?
The federal government has publicly endorsed Stargate and has promoted faster development of U.S. AI infrastructure. The White House lists Stargate as a $500 billion private investment. A White House science-and-technology summary described a plan involving up to 20 large data centers and a $100 billion Abilene flagship project.
Free tools Windows power users keep installed
One-click scans. No signup required.
In July 2025, a federal executive order defined qualifying data-center projects as facilities requiring more than 100 megawatts of new AI-related load and at least $500 million in capital expenditures. It directed agencies to identify possible financing support and other measures to accelerate construction.
That policy involvement does not make Stargate a federal spending program. Political support, permitting policy, tax incentives, loans, guarantees, grants, or other support for qualifying infrastructure are different from a direct $500 billion appropriation to Stargate. The public materials do not establish that the federal government has provided the entire headline amount.
Why OpenAI wants Stargate
- Compute availability: Frontier models require large clusters of accelerators for training.
- Inference demand: Running models for users creates a continuing workload after training is complete.
- Geographic resilience: Multiple campuses reduce dependence on one data center or utility region.
- Supply-chain control: U.S.-based infrastructure can support domestic chip, construction, energy, and manufacturing ecosystems.
- Commercial capacity: Additional infrastructure can support consumer products, enterprise customers, developers, and government users.
- National strategy: OpenAI and policymakers frame large-scale compute as part of international AI competition.
Stargate capacity is not automatically dedicated exclusively to ChatGPT users. The mix of OpenAI workloads, enterprise services, cloud capacity, and other uses depends on the relevant operating and commercial agreements.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.The biggest financial and business risks
Financing uncertainty
The founding announcement named funders and roles but did not disclose the full split between equity, debt, leases, equipment financing, project finance, and corporate commitments. The $500 billion figure therefore does not tell readers who bears the risk if construction costs rise, demand weakens, or a site is delayed.
Overbuilding and hardware obsolescence
Buildings, power connections, and transmission assets can last for decades. GPUs, networking equipment, and server configurations may have much shorter economic lives. A campus built for one accelerator generation may require costly upgrades before the facility itself is fully depreciated.
Demand and utilization
Stargate’s economics depend on sustained demand for AI training and inference. If model economics, customer demand, or usage patterns change, the partners may not use every planned megawatt at the expected rate.
Utility and ratepayer exposure
Promises that a developer will fund infrastructure are meaningful, but readers should still ask whether the utility has approved the arrangement, whether special rates apply, and who pays for stranded assets if the project is canceled or materially downsized.
Concentration risk
The initiative concentrates strategic capacity among a small group of firms, including OpenAI, Microsoft, Oracle, NVIDIA, SoftBank, and specialized infrastructure partners. That can improve coordination while also increasing dependence on a limited number of hardware, cloud, capital, and operating providers.
Recommended Free Tools
What Stargate means for ordinary consumers and investors
Consumers cannot buy a direct ownership share in “Stargate” simply by signing up for an AI service. Investors also should not treat the $500 billion headline as a guaranteed revenue figure for every participant. The commercial benefit is distributed across model providers, cloud companies, chip suppliers, utilities, construction firms, equipment makers, landlords, lenders, and energy developers.
Best Value
- Family farms not data design for people against AI server farms, data center expansion, rural land buyouts, corporate agriculture, and industrial tech development replacing farmland and open space. Rural conservation and anti data center message.
- AI protest design for farmers, land conservation supporters, anti AI activists, sustainability groups, environmental advocates, rural communities, and people opposing server farm construction, power grid strain, and farmland destruction.
- Lightweight, Classic fit, Double-needle sleeve and bottom hem
For anyone evaluating companies exposed to the buildout, the useful questions are:
- Is the company receiving an actual order, lease, or contract, or merely participating in an announcement?
- Is reported capacity planned, under construction, powered, or operational?
- Who owns the equipment and who carries depreciation risk?
- Is revenue recognized from construction, equipment delivery, cloud usage, leasing, or long-term commitments?
- Who pays for power, transmission, cooling, and upgrades?
- Can the infrastructure be repurposed if AI demand or hardware requirements change?
Readers seeking AI capacity rather than investment exposure should choose according to scale: public cloud or specialized GPU cloud for occasional workloads, Azure or OCI for enterprise deployment, negotiated reservations for large training runs, wholesale colocation for physical capacity, and model APIs for users who do not need to own infrastructure. GPU hardware alone is not a complete data-center solution; it also requires power, cooling, networking, support, and skilled operations.
How to interpret the latest capacity claims
OpenAI’s reported figures changed from 10 gigawatts as an original target, to nearly 7 gigawatts of planned capacity in September 2025, to more than 8 gigawatts in October 2025, and then to a claim in April 2026 that the initial 10-gigawatt milestone had been surpassed.
These numbers are not necessarily contradictory. They may describe different dates, projects, contractual stages, and definitions. A careful reader should separate:
- Announced target
- Planned capacity
- Capacity under development
- Leased capacity
- Grid-connected capacity
- Powered capacity
- Operational capacity
- Capacity actively used for training or inference
Public company announcements establish that a substantial buildout is being pursued. They do not provide a single independent accounting that reconciles every site, dollar, megawatt, lease, and operating workload.
Separate treatment of Stargate UAE
OpenAI has also described Stargate UAE as an international deployment involving G42, Oracle, NVIDIA, Cisco, and SoftBank. That project belongs to the broader Stargate platform, but it should not automatically be added to the original U.S. $500 billion investment target.
Bottom line
Project Stargate is best understood as a large, privately financed AI-infrastructure strategy rather than a government project or a single construction site. Its $500 billion figure is a real announced four-year target, but the public record does not show that $500 billion has already been spent, independently audited, or converted into fully operational computing capacity.
The decisive issues are execution and economics: securing power, completing campuses, delivering hardware, controlling cooling and water impacts, financing long-lived infrastructure, and ensuring that AI demand remains strong enough to use it. OpenAI’s reported capacity milestones show an ambitious and expanding buildout; they should still be labeled as company claims and separated from independently verified spending, completion, and utilization.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
