OpenAI’s “Stargate” Project: Powering the Future of AI with Massive Data Centers

OpenAI recently unveiled its flagship Stargate AI data center in Abilene, Texas, confirming that this massive facility is the first of six planned under its ambitious AI infrastructure project. The move—executed in partnership with Oracle and SoftBank—marks a bold shift in OpenAI’s approach to scaling compute capability beyond its previous reliance on a single major cloud provider.

With the Texas site already underway, OpenAI and its partners have announced plans to build five additional data center complexes across other U.S. locations, aiming to reach nearly 7 gigawatts of AI compute capacity in the near term and ultimately fulfilling a $500 billion, 10-gigawatt commitment under “Project Stargate.”

What is Project Stargate?

• Stargate LLC is a joint venture between OpenAI, Oracle, SoftBank, and investment firm MGX, created to build AI compute infrastructure in the U.S.
• The project was formally announced in January 2025 during an event at the White House.
• Its goal: to invest up to $500 billion over several years to deploy 10 gigawatts of AI computing capacity across multiple data centre campuses.
• OpenAI, Oracle, and SoftBank have each committed roles—OpenAI often focusing on the operational / algorithmic side, Oracle on infrastructure, and SoftBank contributing energy design, funding, and regional expertise.

In effect, Stargate is positioned as OpenAI’s multi-region backbone for AI compute, decoupled from cloud provider constraints and capable of scaling at national levels.

The Texas Flagship: Abilene, Texas

Design & Scale

• The Abilene campus is composed of eight “H-shaped” buildings, each designed to house tens of thousands of GPUs (including NVIDIA GB200 chip racks) and to support high-density AI workloads.
• Early phases of the campus are already operational, with infrastructure being brought online in phases.
• The site consumes a significant amount of power—900 megawatts are expected to be drawn from a combination of a new gas-fired plant and regional wind/solar sources.
• To address environmental concerns (particularly in drought-sensitive West Texas), the facility also employs a closed-loop water cooling system to reduce local water usage.

Partnerships & Chip Supply

• Oracle is committing to purchasing and leasing hundreds of thousands of NVIDIA GB200 GPUs to support the compute operations in Abilene and beyond.
• Reports suggest Oracle and OpenAI will develop an additional 4.5 GW of capacity beyond the Abilene campus as part of the Stargate roadmap.
• The Abilene site is already being used for early training and inference workloads, indicating that the facility isn’t just a “build-to-come” project—it’s already contributing to OpenAI’s compute needs.

Five New Sites: Where and Why

OpenAI, Oracle, and SoftBank have chosen five new locations for upcoming Stargate campuses, bringing geographical diversity and growth capacity.

Key Locations

• Shackelford County, Texas – near the Abilene campus, built in partnership with Oracle.
• Doña Ana County, New Mexico – a southwestern expansion under Oracle’s lead.
• Midwest (undisclosed) – a site to be announced, likely to provide central U.S. coverage.
• Lordstown, Ohio – under SoftBank / SB Energy’s purview.
• Milam County, Texas – a SoftBank-led Texas expansion beyond Abilene.

These new sites, combined with a 600 MW expansion at Abilene, bring total planned Stargate capacity to nearly 7 gigawatts, with the project on track to meet or exceed its $500 billion / 10 GW goal.

OpenAI states that the five new sites involved a selection process reviewing over 300 proposals from 30+ states.

Expected job creation is significant: 25,000 onsite roles and many more indirectly via construction, logistics, services, and supply chain.

Strategic Implications & Challenges

Strategic Upsides

1. Independence from a single cloud provider – Stargate gives OpenAI more control over infrastructure, reducing dependencies on third-party cloud contracts.
2. Scale for next-gen models – Massive compute growth is necessary to train and serve increasingly large AI models; Stargate is intended to scale that foundation.
3. Geographic diversity & latency – Multiple campuses spread across U.S. regions help reduce latency and provide redundancy.
4. Energy integration & sustainability – Incorporating renewable sources, efficient cooling, and local energy planning to mitigate environmental footprint.

Key Challenges

• Power & energy constraints – Procuring enough clean or stable power to feed multi-gigawatt sites is nontrivial, especially in regions with limited grid capacity.
• Cost & capital risk – A $500 billion commitment is enormous; justifying returns and avoiding stranded capacity will be critical.
• Regulatory and environmental scrutiny – Large energy and water usage, local land impact, and community concerns may draw opposition.
• Chip supply & logistic bottlenecks – Ensuring steady delivery of high-end GPUs, networking, cooling, and infrastructure hardware at scale is nontrivial.
• Operational complexity & integration – Managing and harmonizing multiple large campuses, software, networking, failover, and security is a massive engineering challenge.

Conclusion

OpenAI’s Stargate project is arguably one of the most audacious AI infrastructure plays in recent years. With the Texas hub now in motion and five more centres in the pipeline, the push reflects how compute scale has become inseparable from AI capability. Through partnerships with Oracle and SoftBank, OpenAI is reinventing how AI compute will be owned, managed, and expanded at national scale.

The next few years will test whether the ambition lives up to the engineering, financial, and regulatory complexities it must overcome. If successful, Stargate may set a blueprint not just for AI deployment in the U.S., but globally for how massive compute can be built and governed.