OpenAI Stargate Project: Scope, Partners, and Evidence

OpenAI Stargate AI Infrastructure Project

Stargate is OpenAI's long-term infrastructure program for securing and building compute capacity with financial, cloud, chip, data-center, energy, construction, and community partners.

The name refers to a multi-party infrastructure effort, not one completed data center or a single funding round. Planned investment, secured power capacity, deployed sites, utilized compute, and accounting cost must be tracked separately.

Initial announcement

OpenAI announced The Stargate Project on January 21, 2025. The announcement described a new company that intended to invest $500 billion over four years in United States AI infrastructure and begin deploying $100 billion.

OpenAI named SoftBank, OpenAI, Oracle, and MGX as initial equity funders. It named Arm, Microsoft, NVIDIA, Oracle, and OpenAI as initial technology partners. SoftBank was assigned financial responsibility and OpenAI operational responsibility in the announcement.

The wording matters. "Intends to invest" is a plan. It should not be reported as $500 billion already spent, funded, or placed into service.

flowchart LR
    A["Financing and partner agreements"] --> B["Sites, power, permits, chips, and construction"]
    B --> C["Secured capacity"]
    C --> D["Deployed and available compute"]
    D --> E["Training and inference utilization"]
    E --> F["Models, products, demand, and revenue"]

A milestone at one stage does not prove completion of the next.

2026 capacity update

On April 29, 2026, OpenAI published a Stargate infrastructure update. It said the original goal was to secure ten gigawatts of United States AI infrastructure by 2029 and that the company had surpassed the milestone, with more than three gigawatts added in the prior ninety days.

The statement is a first-party report of secured infrastructure. It does not provide a site-by-site audited schedule of deployed, energized, available, contracted, reserved, or utilized capacity.

The same post describes Abilene, Texas as a flagship site and says OpenAI trained GPT-5.5 there using Oracle Cloud Infrastructure and NVIDIA GB200 systems. Product and model claims should be checked again at publication because the infrastructure and model record changes quickly.

Relationship to OpenAI's broader compute strategy

OpenAI's March 2026 financing announcement describes a broader portfolio across Microsoft, Oracle, AWS, CoreWeave, Google Cloud, NVIDIA, AMD, AWS Trainium, Cerebras, Broadcom, and data-center partners.

Stargate is therefore one organizing infrastructure program inside a multi-provider strategy. It should not be described as OpenAI owning every layer outright or depending on one cloud.

OpenAI presents compute as a flywheel that supports research, products, adoption, revenue, and reinvestment. That is the company's strategic thesis. Public sources do not independently establish utilization, cost per unit, operating leverage, or return on the program.

What to track

Future updates should retain the date, source, unit, geographic scope, partners, status, and difference between capacity and use.

Evidence boundary

OpenAI's pages establish the project's announced structure, named partners, intended investment, current company-reported capacity, and strategic framing. They do not provide a complete audited capital ledger, utilization record, power draw, schedule, financing waterfall, or economic return.

[[How to Analyze AI Lab Economics]] provides the metric discipline needed to avoid converting infrastructure ambition into unsupported financial conclusions.

This profile reflects current OpenAI sources reviewed on July 27, 2026. AI assistance was used for research organization, drafting, and validation. Publication remains unauthorized.