Stargate Project

Stargate is a large-scale artificial intelligence infrastructure programme announced in January 2025 by OpenAI, SoftBank, Oracle and the Abu Dhabi-backed investment group MGX. The partners describe it as an overarching AI infrastructure platform intended to deploy gigawatt-scale data centre campuses for training and serving frontier AI models, with a stated commitment of US$500 billion of investment over four years and capacity of up to 10 gigawatts (GW).^[1]

History

Stargate was unveiled at the White House on 21 January 2025, with initial equity funding from OpenAI, SoftBank, Oracle and MGX, SoftBank chairman Masayoshi Son as chairman, and OpenAI responsible for operations. The first flagship campus was sited on the outskirts of Abilene, Texas, on a 1,100-acre property, with Oracle building and operating the facilities on its cloud infrastructure.^[1]

Construction moved unusually quickly. Oracle began delivering the first NVIDIA GB200 racks in June 2025, and two of the campus's eight planned buildings were operational by September 2025, running early training and inference workloads for OpenAI's next-generation research.^[2] In September 2025, the partners announced five additional United States sites - in Shackelford County, Texas; Doña Ana County, New Mexico; Port Washington, Wisconsin; Lordstown, Ohio; and Milam County, Texas - taking the programme to nearly 7 GW of planned capacity and more than US$400 billion of intended investment over three years.^[2][3]

Related international projects followed, including Stargate UAE, announced in 2025 with Abu Dhabi's G42 group and partners including Oracle, NVIDIA, Cisco and SoftBank.^[6] The Stargate Norway campus, announced with Nscale and Aker, changed hands in April 2026 when Microsoft took over the capacity after OpenAI did not conclude an offtake agreement.^[5]

Key concepts

Stargate treats compute, power and capital as a single integrated system. Industry analysts describe the model as a shift toward vertically coordinated "AI industrial parks" rather than conventional hyperscale data centres: campuses combine behind-the-meter power generation, co-located energy infrastructure, liquid cooling and phased GPU deployment, with total capacity measured in gigawatts.^[4]

At full build-out the Abilene campus alone is planned to draw about 1.2 GW and to host hundreds of thousands of NVIDIA GPUs, making it among the largest single AI computing sites in the world.^[2] SoftBank acts as the principal financial backer, Oracle supplies cloud infrastructure at several campuses, and SB Energy, a SoftBank subsidiary, provides powered infrastructure for fast-build sites in Texas.^[2][4]

Applications and impact

Stargate is the largest single commitment in a broader wave of AI data centre construction, and it has become a reference point in debates about compute sovereignty: governments in the Gulf, Europe and Asia have pursued similar sovereign AI partnerships, and the vendor alliance around the project spans NVIDIA, Microsoft and Arm.^[4] The partners have claimed more than 25,000 onsite jobs across the new sites, although independent reporting has questioned how much permanent operational employment such campuses ultimately create.^[2]

The build-out has also attracted scrutiny over energy demand, water use, grid interconnection and community consent. In 2026 the partners adjusted parts of the plan, scaling back a planned expansion at Abilene and reassigning capacity, and analysts described financing and demand certainty as the programme's central open questions.^[4]

>See Also

• OpenAI
• NVIDIA
• AI Data Centres in Malaysia
• YTL AI Cloud
• Sovereign AI
• AI Chip Export Controls