OpenAI Infrastructure Tracker
Mapping the Global Expansion of Project Stargate and AI Compute Partnerships
Last updated: November 2025 | Version 1.4 (Updated with SB OAI Japan)
OpenAI is executing one of the largest private-sector infrastructure programs in history, targeting more than 10 GW of power capacity across purpose-built AI campuses known as Project Stargate as well as long-term GPU cloud partnerships.
This tracker consolidates verified sites, partners, and deals — in the United States and abroad — associated with OpenAI’s compute build-out, including projects under construction, announced international expansions, and major colocation or capacity-procurement contracts.
The data below is drawn from corporate filings, municipal and environmental records, and official press releases from OpenAI, Oracle, Microsoft, AWS, CoreWeave, SoftBank, and regional development agencies. Converge Digest will continue updating this tracker as new Stargate phases, GPU-capacity agreements, and AI ventures emerge.
OpenAI’s infrastructure effort blends direct build-outs and long-term cloud commitments. In the U.S., multiple Project Stargate sites are already in motion, representing over 8 GW of capacity. Internationally, new sites in the UAE, Norway, and Argentina mark the first phase of a global expansion supported by Oracle, SoftBank, and regional energy partners.
Beyond physical campuses, OpenAI continues to lease compute capacity through multibillion-dollar agreements with Microsoft (Azure), Oracle Cloud Infrastructure, AWS, and CoreWeave. The strategy provides both control over custom-built AI factories and flexible access to hyperscale GPU clouds. Following OpenAI’s 2025 reorganization — splitting its non-profit foundation and public-benefit corporation — the company gained greater flexibility to raise capital and form international joint ventures such as SB OAI Japan.
Direct Builds – United States
| Project | Location | Capacity (GW) | Partners / Developers | Status / Completion |
|---|---|---|---|---|
| Stargate 1 | Abilene, TX | ≈1.2 (planned) | OpenAI, Oracle, Lancium, Crusoe, Blue Owl, Primary Digital Infrastructure | Under construction; first phases live 2025 |
| Frontier | Shackelford County, TX | 1.4 (planned) | OpenAI, Oracle, Vantage Data Centers, INNIO / Voltagrid | Early construction; target 2026+ |
| Lighthouse | Port Washington, WI | ≈1.0 (planned) | OpenAI, Oracle, Vantage Data Centers, DBRG, WEC Energy | Planning; completion target 2028 |
| Project Jupiter | Doña Ana County, NM | Up to 1.5 | OpenAI, STACK Infrastructure, BorderPlex Digital, Blue Owl | Permitting; multi-phase build 2025–27+ |
| Lordstown | Lordstown, OH | TBD (part of 7-GW U.S. expansion) | OpenAI, SoftBank | Announced; site work 2025–26 |
| Milam County | Milam County, TX | TBD (part of 7-GW U.S. expansion) | OpenAI, SoftBank | Announced; fast-build site 2026+ |
| The Barn | Saline Township, MI | >1.0 (planned) | OpenAI, Oracle, Related Digital, DTE Energy | Announced; construction starts 2026 |
International Stargate Sites
| Project | Location | Capacity (GW) | Partners / Developers | Status / Completion |
|---|---|---|---|---|
| Stargate UAE | Abu Dhabi, UAE | 1.0 (phase 1 of 5-GW campus) | G42 / Khazna, OpenAI, Oracle, NVIDIA, Cisco, SoftBank | Under construction; first 200 MW live 2026 |
| Stargate Norway | Kvandal / Narvik, Norway | 0.23 initial (up to 0.52) | Nscale & Aker (50/50 JV), OpenAI anchor | Announced; initial phase 2026, 100k NVIDIA GPUs |
| Stargate Argentina | Patagonia, Argentina | Up to 0.5 (planned) | Sur Energy (developer), OpenAI anchor customer | LOI signed; planning underway, post-2027 |
Capacity Procurement & Colocation Partnerships
| Partner | Region | Estimated Capacity | Deal Type | Status |
|---|---|---|---|---|
| Microsoft (Azure) | Global | TBD ($250B+ Azure commit) | Long-term GPU & cloud capacity contract | Active |
| Oracle Cloud Infrastructure | U.S. (4.5-GW Stargate) + global OCI | 4.5 GW (Stargate tranche) | Colocation + joint-build data center shells | Active |
| AWS | Global | TBD (7-year, $38B) | Cloud capacity deal for GB200/GB300 clusters | Active |
| CoreWeave | U.S. & EU | TBD ($22.4B contracts) | Multi-deal GPU cloud (Training + Inference) | Active |
| Lambda | U.S. | TBD | GPU cloud provider (OpenAI listed customer) | Active |
| SoftBank (Japan) / SB OAI Japan | Japan (enterprise AI JV) | TBD | 50-50 joint venture for enterprise AI solutions (“Crystal Intelligence”); includes planned infrastructure investment | Active; launch 2026 |
Legend & Methodology
≈ denotes approximate capacity based on developer statements or filings. Up to refers to maximum design load at full build-out. TBD means data not publicly disclosed or still under negotiation.
Figures are derived from official press releases, corporate disclosures, municipal permits, and industry filings through November 2025. Capacities are rounded to the nearest 0.1 GW. Timelines represent earliest announced milestones and may shift as projects progress.
The OpenAI Infrastructure Tracker will be updated as new Stargate sites, PPAs, or GPU-capacity contracts emerge. Send updates to [email protected].
🌐 Follow our ongoing coverage at https://convergedigest.com/category/ai-infrastructure/
