Introduction

The backbone of AI progress is compute power. This essential truth underpins OpenAI’s Stargate initiative, which recently reached a significant milestone: OpenAI, Oracle, and SoftBank have announced five new data center sites across the U.S. These additions are set to accelerate the development of a sophisticated AI infrastructure. The partners are confident that Stargate is on track to fulfill its ambitious commitment of $500 billion and 10 gigawatts of capacity ahead of schedule, transforming a visionary plan into a tangible reality, complete with jobs and advanced capabilities.

In this article, we will explore the essence of Stargate, the significance of these new sites, and how a collaborative ecosystem is propelling the next wave of AI advancement.

What is Stargate?

Stargate is OpenAI’s long-range infrastructure platform and investment initiative aimed at providing the compute, power, and connectivity essential for next-gen AI systems. Announced in January 2025, Stargate plans to invest up to $500 billion over four years to deploy 10 gigawatts of AI capacity across the United States. This initiative has received backing from key industry partners and the White House.

Stargate operates through a consortium model: OpenAI establishes the research and operational requirements, while partners like Oracle, SoftBank, and CoreWeave contribute their expertise in designing, constructing, and operating the data centers at a scale never seen before. This collaborative approach enables faster progress compared to traditional construction methods.

Overview of the New Sites

The newly announced data center sites extend Stargate’s reach across several U.S. regions:

• Shackelford County, Texas (OpenAI and Oracle)
• Doña Ana County, New Mexico (OpenAI and Oracle)
• A to-be-announced location in the Midwest (OpenAI and Oracle)
• Lordstown, Ohio (OpenAI and SoftBank)
• Milam County, Texas (OpenAI and SoftBank, in collaboration with SB Energy)

With the flagship campus in Abilene, Texas, and other ongoing projects, these selections aim to bring Stargate to nearly 7 gigawatts of planned capacity and over $400 billion in investments within the next three years. The sites were carefully chosen from over 300 proposals submitted by more than 30 states.

Capacity, Timing, and Job Creation

• Oracle-Developed Sites: The trio of new sites (Shackelford County, TX; Doña Ana County, NM; and the Midwest location) and an additional potential expansion near Abilene are projected to deliver over 5.5 gigawatts of capacity.
• SoftBank-Developed Sites: Lordstown, OH and Milam County, TX are designed to scale to 1.5 gigawatts in approximately 18 months.
• Employment Opportunities: The overall initiative is expected to generate over 25,000 onsite jobs, alongside tens of thousands more across supply chains and local services as construction progresses.

These figures illustrate the significance of the five-site announcement: it transitions Stargate from concept to actionable execution at multi-gigawatt capacity, laying the groundwork for next-generation AI model training and inference.

Significance for AI Advancement

As large AI models evolve, so does the demand for reliable, cost-effective, and low-latency compute resources. Stargate is strategically positioned to tackle this demand by establishing a comprehensive blueprint for creating and connecting AI data centers at a national scale. This will result in quicker research iterations, reduced times for deploying new features, and enhanced access to high-performance compute resources for developers and enterprises.

Oracle’s Role: OCI Capacity and Rapid Scaling

A cornerstone of this project is Oracle Cloud Infrastructure (OCI). In July 2025, OpenAI and Oracle signed an agreement to develop 4.5 gigawatts of additional Stargate capacity in the U.S., building on the existing Abilene campus that has started processing early training and inference workloads. Oracle began supplying NVIDIA GB200 racks in June to enhance OpenAI’s upcoming research efforts.

Oracle and NVIDIA have been enhancing their collaboration, making the NVIDIA GB200 NVL72 systems available on OCI Supercluster. This facilitates large-scale training with advanced GPUs and high-speed networking engineered specifically for AI applications, aligning with Oracle’s plan to implement large-scale AI infrastructure swiftly and affordably in distributed cloud regions.

SoftBank’s Contribution: Focused on Fast-Build Designs

SoftBank’s two sites add a focus on quickly built, high-density data centers with fully powered infrastructure upon launch. The Lordstown, Ohio campus is under construction and on track to be operational next year, while Milam County, Texas is being developed in collaboration with SB Energy, a SoftBank Group company dedicated to enabling rapid energy deployment. Together, these sites can scale to 1.5 gigawatts in about 18 months, helping diversify Stargate’s geographical and supplier landscape.

Energy and Sustainability Considerations

Data centers at a gigawatt scale require substantial and dependable power. Consequently, site selection prioritizes access to transmission capacity, quick interconnect capabilities, and opportunities for co-locating with new energy generation and storage options. Partnerships with utilities and energy developers will be vital to introducing new power capacity without incurring unpredictable costs.

For instance, in Texas, the Milam County site integrates AI infrastructure with SB Energy’s power expertise, with local news indicating progress on substantial solar projects in the region. This reflects an industry trend towards combining clean energy sources with high-performance compute resources.

How the Pieces Fit Together

• Location Strategy: A strategic mix of Texas, New Mexico, Ohio, and a future Midwest site helps spread risk across different grids and climates, positioning capacity near major fiber routes.
• Modular Designs: Standardized structures for power, cooling, and networking shorten the time from groundbreaking to operational capacity.
• Supply Chain Partnerships: Coordinated commitments from chip manufacturers, power supply organizations, construction firms, and networking vendors aim to alleviate potential bottlenecks.
• Cloud Integration: Utilizing OCI alongside other partner ecosystems allows OpenAI to optimize workload distribution based on performance needs.

Ahead of Schedule on the Original Commitment

OpenAI, Oracle, and SoftBank assert that the five new sites, along with efforts at the Abilene campus and collaboration with other partners, put Stargate firmly on track to meet its full $500 billion and 10-gigawatt commitment by the end of 2025. This commitment was initially announced by the White House when Stargate was first unveiled in January 2025.

The NVIDIA Factor: 10 GW Systems on the Horizon

In parallel, OpenAI and NVIDIA have inked a letter of intent to deploy a minimum of 10 gigawatts of NVIDIA systems, backed by NVIDIA’s plan to invest up to $100 billion as these systems become operational. The first gigawatt of systems is expected to be available in the second half of 2026, aligning with Stargate’s timeline and construction goals.

A Closer Look at Each Cluster

• Abilene, Texas: Already operational, providing early training and inference with plans for 600 megawatts of additional capacity nearby.
• Shackelford County, Texas: Part of Oracle’s trio, this site is intended for fast scaling and interconnection with Texas’s broader grid and fiber systems.
• Doña Ana County, New Mexico: Offers geographic diversity in the Southwest and is positioned near crucial long-distance routes in collaboration with Oracle.
• Midwest Site (TBA): The future Oracle-led site, strategically located to reduce latency while tapping into regional power markets.
• Lordstown, Ohio: Employs an advanced design aimed to be operational next year, reinforcing the PJM grid region.
• Milam County, Texas: Developed in collaboration with SB Energy, this site is designed for rapid infrastructure deployment.

Implications for Developers and Enterprises

• Increased Capacity: The additional resources will reduce wait times for significant training tasks, facilitating larger context windows and more intricate multi-agent processes.
• Lower Latency and Higher Availability: The geographical diversity of sites enhances service proximity to users and provides regional failover options.
• Cost and Efficiency Improvements: Standardized builds, contemporary power distribution methods, and high-utilization clusters are expected to improve cost-effectiveness over time.
• Accelerated Innovation: Enhanced compute availability will enable research teams to conduct more experiments simultaneously, shorten training periods, and roll out updates more quickly.

Risks and Areas for Attention

• Power and Grid Integration: The pace at which utilities and regional organizations can approve and deliver necessary interconnects for multi-hundred-megawatt developments remains a concern.
• Supply Chain Resilience: Timelines for acquiring advanced GPUs and networking gear are still tight; coordinating across partners is essential.
• Local Impact: Communities will consider the implications of water and land use, alongside the potential for job creation and tax revenue.
• Policy Stability: The success of multiyear investments hinges on consistent permitting, incentives, and policies that may outlast political cycles.

Understanding Scale

It is beneficial to separate two aspects of scale:
– Electrical Capacity (Gigawatts): This refers to the power required to operate clusters at maximum capacity. Gigawatts indicate the number of high-density systems that can run concurrently.
– Compute Capacity (Chips and Systems): This encompasses the quantity and types of GPUs, CPUs, interconnects, and storage. For example, NVL72-class systems compile numerous GPUs per rack to expedite training; interlinking thousands of these forms the superclusters used for cutting-edge models.

Timeline Snapshot

• January 2025: Stargate is announced with a commitment of $500 billion and 10 gigawatts.
• June 2025: Oracle starts delivering NVIDIA GB200 racks for initial workloads in Abilene.
• July 2025: Agreement signed for an additional 4.5 GW of company capacity in the U.S.
• September 2025: Announcement of five new sites, pushing planned capacity to nearly 7 GW and expected investments over $400 billion in the next three years.
• September 2025: OpenAI and NVIDIA affirm plans for 10 GW of systems, with the first gigawatt aimed for H2 2026.

Success Metrics

If Stargate meets its defined milestones, the United States will see the rapid addition of multiple gigawatts of specifically designed AI capacity. This will provide researchers and developers with faster access to compute resources, more reliable performance, and the capability to address new challenges previously limited by infrastructure.

For the communities hosting these data centers, success will depend on how effectively the partners collaborate with local leaders to create jobs and increase tax revenues while managing environmental, water, and grid impact concerns. Early indications suggest a focus on speed and scalability, but the overall effectiveness and community engagement will ultimately determine long-term outcomes.

Key Takeaways

• Five new data center sites enhance Stargate’s U.S. presence and increase planned capacity to nearly 7 GW.
• Oracle’s OCI will supply a substantial portion of this capacity, with GB200-class systems currently arriving in Abilene.
• SoftBank’s Ohio and Texas sites focus on rapid deployment with powered infrastructure.
• The program stays on track to achieve a total commitment of $500 billion and 10 gigawatts by the end of 2025, in alignment with NVIDIA’s broader roadmap.

FAQs

Q1: What exactly is Stargate?

Stargate is OpenAI’s infrastructure platform aimed at building up to 10 gigawatts of AI data center capacity in the U.S., supported by an investment of up to $500 billion over four years, coordinating sites, power, chips, and networking through partnerships with companies like Oracle and SoftBank.

Q2: Where are the five new sites located?

The new sites are in Shackelford County, TX; Doña Ana County, NM; a future location in the Midwest; Lordstown, OH; and Milam County, TX, complementing the main campus in Abilene, TX.

Q3: When will these sites become operational?

Timelines vary by site. The Lordstown campus is slated to launch next year, with both SoftBank sites expected to reach 1.5 GW of capacity in around 18 months. Oracle-led sites will commence operations as equipment and power obtain necessary interconnections.

Q4: How does this relate to NVIDIA’s plans?

OpenAI and NVIDIA have signed a letter of intent to deploy at least 10 GW of NVIDIA systems, beginning with 1 GW planned for the second half of 2026. This aligns chip availability with Stargate’s construction schedules.

Q5: What about sustainability?

Each site will require thorough power planning, and partners are aligning AI campuses with new energy generation and grid investments. In Texas, development is linked with SB Energy’s powered infrastructure and regional renewable initiatives. Details will evolve as interconnections and power agreements are finalized.

Conclusion

The expansion of Stargate with five new data center sites represents a critical milestone in establishing the compute foundation required for future AI breakthroughs. It signifies a remarkable blend of vision, financing, and execution at a national scale. If the partners stay on course with their commitments related to power supply, supply chain management, and effective community engagement, the result will be not just larger computing clusters, but a sustainable platform for innovation that brings benefits to researchers, developers, and users across the economy.