Status: Confirmed / Under Rapid Construction | Location: Multiple U.S. states, including Texas (Abilene, Shackelford, Milam), New Mexico, Ohio (Lordstown), and Wisconsin (Midwest site)
The Stargate AI Infrastructure Platform is a landmark data center initiative driven by a historic partnership between OpenAI, Oracle, and SoftBank, representing a $500 billion USD commitment to build the computational backbone for the next era of Artificial Intelligence. This project is not a single data center but an interconnected platform of AI-Native Gigascale campuses across the United States, targeting a colossal total capacity of 10 Gigawatts (GW) by 2030. The confirmed expansion announced in late 2025 brings the secured capacity to nearly 7 GW with an estimated investment of over $400 billion in the near term.
The core distinction of Stargate is its purpose: it is being built specifically for large-scale AI training and inference. This mandate requires an infrastructure fundamentally different from traditional hyperscale facilities. The facilities must be engineered to house and cool dense racks of NVIDIA GB200 and similar next-generation GPUs, necessitating the universal adoption of liquid cooling systems (such as direct-to-chip or immersion cooling) to manage power densities exceeding 250 kW per rack.
The project’s rapid expansion is focused on securing massive land and power reserves in strategically chosen locations. While the flagship site is already underway in Abilene, Texas, the partnership has quickly confirmed additional sites in New Mexico, Ohio (Lordstown, where SoftBank has broken ground), and is leveraging partnerships, such as with Vantage Data Centers in Shackelford County, Texas, to scale rapidly.
Oracle is the lead cloud provider, hosting the Stargate platform on its Oracle Cloud Infrastructure (OCI), and SoftBank is leveraging its energy expertise (SB Energy) to ensure the power supply, which is the single biggest constraint for such large-scale AI projects. The project is a confirmed, massive capital deployment that underscores the urgent, global race to acquire the compute necessary to drive AI development. The massive scale and focus on dedicated AI hardware make this one of the most significant and confirmed infrastructure projects currently under development globally.
| Field | Value |
|---|---|
| Project Name | OpenAI/Oracle/SoftBank Stargate AI Infrastructure Platform |
| Location | Multiple U.S. states, including Texas (Abilene, Shackelford, Milam), New Mexico, Ohio (Lordstown), and Wisconsin (Midwest site) |
| Status | Confirmed / Under Rapid Construction |
| Commissioning | Initial capacity (Abilene) operational; major capacity by end of 2026 and fully by 2030 |
| Total IT Load | Not specified (Part of multi-GW capacity) |
| Total Capacity | Planned total capacity of 10 GW (Gigawatts) |
| Tier Level | Likely Tier III+ to Tier IV (Mission-critical AI infrastructure) |
| Project Type | AI-Native Gigascale Hyperscale Data Center Campus |
| Header | Details |
|---|---|
| City Name | Abilene, Texas (Flagship) and others |
| Population | Not specified (Rural/Suburban campus locations) |
| Urban Agglomeration | Not specified (Rural/Suburban campus locations) |
| City GDP | Not applicable (Multi-site, national focus) |
| Per Capita Income | Not applicable (Multi-site, national focus) |
| City Tier | Emerging AI Compute Hubs |
| Key Strengths | Access to vast amounts of affordable land; Favorable power supply/grid access; Strategic location for new-generation AI-compute clusters. |
| Header | Details |
|---|---|
| Developer / Operator | Oracle (Cloud Infrastructure/OCI) and SoftBank (Powered Infrastructure) |
| Strategic Infrastructure Partner | OpenAI (Primary tenant/driver), CoreWeave, Vantage Data Centers |
| Construction Contractor | Multiple, including those associated with SoftBank/SB Energy and partners like Vantage |
| MEP Engineering | Designed explicitly for ultra-high-density NVIDIA GB200 GPU racks |
| Network Connectivity | High-speed, dedicated fiber networks linking campuses to support massive data transfer for AI model training |
| Power Infrastructure | Dedicated power generation/purchase agreements; focus on securing multi-GW power supply |
| Header | Details |
|---|---|
| Power Capacity | 10 GW (Target by 2030) |
| UPS Redundancy | Expected highest levels of redundancy (e.g., 2N) due to project criticality |
| Cooling System | Mandatory Liquid Cooling (Direct-to-chip/Immersion) for ultra-high-density AI hardware (e.g., 250 kW+ per rack) |
| Connectivity | Supports OpenAI's large model training and deployment; integrated with Oracle Cloud Infrastructure (OCI) |
| PUE Target | Aggressive low PUE targets, enabled by liquid cooling and optimized power systems |
| Energy Mix | Focus on secure, dedicated power supply, including large-scale renewable energy (PPA) and potentially nuclear (as seen in other Meta/Constellation deals) |
| Header | Details |
|---|---|
| Announcement | Initial $500B commitment announced January 2025; expansion to 7 GW confirmed September 2025 |
| Construction Start | Ongoing (Lordstown, OH site has broken ground; Abilene, TX is operational) |
| Phase 1 Go-Live | Multi-site, multi-gigawatt development over 3-5 years |
| Full Buildout | Expected by 2030 (Full 10 GW capacity) |
| Header | Details |
|---|---|
| Total Investment | $500 Billion USD (Total potential commitment over several years) |
| Funding | OpenAI, Oracle, SoftBank, and various investment partners (e.g., CoreWeave, debt/equity financing) |
Subscribe to our exclusive newsletter and get the latest insights on data center trends, market forecasts, and infrastructure innovations delivered straight to your inbox.