What $500 billion in AI infrastructure actually looks like: inside OpenAI's Stargate

Updated December 11, 2025

December 2025 Update: Project reaching 8GW planned capacity and $450B+ committed investment—ahead of schedule. Abilene flagship at 1.2GW with 400,000 GB200 GPUs planned. SoftBank and OpenAI each contributed $19B (40% ownership). NVIDIA committing $100B in processor purchases. AMD providing GPUs for 6GW future capacity. SMRs planned for baseload power.

President Donald Trump called it "the largest AI infrastructure project in history" when he stood alongside OpenAI CEO Sam Altman, SoftBank CEO Masayoshi Son, and Oracle Chairman Larry Ellison on January 21, 2025.¹ The Stargate Project intends to invest $500 billion over four years building AI infrastructure for OpenAI across the United States. The scale has been compared to the Manhattan Project. By late 2025, the project reached 8 gigawatts of planned capacity and over $450 billion in committed investment, ahead of the original schedule.²

The numbers demand examination. Eight gigawatts equals the electricity consumption of a small nation. The flagship Abilene, Texas campus alone will draw 1.2 gigawatts once all eight AI factory halls come online, enough to power 750,000 homes.³ The project will deploy up to 400,000 NVIDIA GB200 Blackwell GPUs at Abilene alone.⁴ Understanding what this infrastructure actually entails reveals both the ambitions and the challenges of building AI at civilizational scale.

Stargate's structure and ownership

Stargate LLC, incorporated in Delaware, joins OpenAI, SoftBank, Oracle, and investment firm MGX in a joint venture with clearly delineated responsibilities.⁵ SoftBank has financial responsibility and serves as chairman through CEO Masayoshi Son. OpenAI has operational responsibility.

The initial funding reflects the partners' commitment levels. SoftBank and OpenAI each contributed $19 billion and hold 40% ownership interest.⁶ Oracle and MGX each contributed $7 billion. Remaining funds come from limited partners and debt financing. SoftBank sold all of its remaining NVIDIA shares in November 2025 to reallocate $30 billion toward its Stargate investment.⁷

Oracle, NVIDIA, and OpenAI collaborate closely to build and operate the computing systems. NVIDIA signed on heavily, with OpenAI announcing a $100 billion investment from NVIDIA to purchase AI processors for the data centers.⁸ AMD agreed in October 2025 to provide GPUs for 6 gigawatts of future deployment capacity.⁹ The partnership extends OpenAI's collaboration with NVIDIA dating to 2016 and its newer partnership with Oracle.

The Texas flagship and beyond

The flagship facility in Abilene, Texas, known internally as "Project Ludicrous," began receiving NVIDIA GB200 server racks in mid-2025 and operates on Oracle Cloud Infrastructure.¹⁰ The 875-acre site will scale to 1.2 gigawatts across eight AI factory halls. By the end of summer 2025, 16,000 GB200 GPUs were installed, with 64,000 planned by end of 2026.¹¹

Five additional sites announced in September 2025 brought the project to nearly 7 gigawatts of planned capacity: Shackelford County, Texas; Doña Ana County, New Mexico; Lordstown, Ohio; Milam County, Texas; and Wisconsin.¹² The Oracle-built campuses plus a 600-megawatt expansion at Abilene deliver approximately 5.5 gigawatts combined.¹³ Two SoftBank campuses in Lordstown and Milam County can scale to 1.5 gigawatts within 18 months.¹⁴

The architecture reflects lessons learned from hyperscale data center development. Industry observers expect the sites to achieve extreme AI density in the range of 100-150 kilowatts per rack.¹⁵ At those power densities, each facility hosts tens of thousands of racks dedicated to training and inference workloads. Direct-to-chip liquid cooling within closed-loop systems enables the high rack densities that AI training demands.¹⁶

Power solutions at unprecedented scale

Stargate's energy requirements create challenges no previous data center project has faced. The project plans to deploy small modular nuclear reactors to establish stable baseload power, with advanced reactors positioned near demand centers to minimize transmission losses.¹⁷

The West Texas region offers abundant renewable energy that the project leverages through integrated power solutions. Large on-site battery storage handles renewable intermittency. Dedicated solar farms serve daytime loads. Backup power comes from 360 megawatts of on-site natural gas turbines.¹⁸ The multi-source approach ensures reliability while capturing renewable economics.

The power requirements strain grid infrastructure. Eight gigawatts of new load requires generation, transmission, and distribution capacity that may not exist in all target locations. Grid interconnection timelines can extend for years. The project must navigate utility negotiations, regulatory approvals, and physical construction to bring power to each site.

Global expansion under OpenAI for Countries

Stargate extended internationally through the OpenAI for Countries initiative. Stargate UAE, the first international project, was announced on May 22, 2025, developing a 1-gigawatt AI infrastructure cluster in Abu Dhabi.¹⁹ NVIDIA, Cisco, and OpenAI partnered with G42, Oracle, and SoftBank to build the facility, expected to open in 2026.

Stargate Norway launched on July 31, 2025, the first European deployment featuring AI data center infrastructure powered by renewable hydropower.²⁰ The UK is developing sovereign AI compute capacity through a partnership with NVIDIA and Nscale. Stargate Argentina, announced October 10, 2025, represents an estimated $25 billion investment in Patagonia and marks the first Latin American site.²¹

The international expansion demonstrates the global demand for AI compute capacity. Each country seeks sovereign AI infrastructure to support domestic AI development while meeting data residency requirements. OpenAI positions itself as the partner enabling that sovereignty while building its global inference network.

Challenges and skepticism

The project faced skepticism from the start. Bloomberg reported in August 2025 that the project had not started and no funds were raised to meet the initial budget.²² Market uncertainty, American trade policy, and AI hardware valuations caused delays according to the report.

The September progress announcements addressed some concerns. The 8 gigawatt capacity and $450 billion committed investment demonstrated momentum beyond the initial announcement. The Abilene facility operating with Oracle Cloud Infrastructure proved the project could deliver functioning data centers, not just press releases.

Questions remain about the full $500 billion. Financing that scale requires debt markets willing to lend against AI infrastructure with uncertain revenue projections. Partner commitments must convert to actual capital flows. Construction must proceed on timelines that match financing windows. The project operates at the edge of what private capital markets can support.

What Stargate means for AI infrastructure

Stargate establishes a new scale for AI infrastructure. Previous projects measured in hundreds of megawatts; Stargate measures in gigawatts. The shift from hyperscale data centers to AI factories reflects the compute intensity that frontier model training demands. Each generation of AI models requires more compute. Stargate builds for the models OpenAI will train in 2027 and 2028.

The project validates the thesis that AI infrastructure constitutes strategic national infrastructure. The White House announcement, the distributed manufacturing locations, the emphasis on American jobs and investment all position AI data centers alongside semiconductor fabs as essential to national competitiveness. Other nations responded with their own sovereign AI infrastructure initiatives, often in partnership with OpenAI through the Countries program.

For organizations planning AI strategies, Stargate demonstrates the infrastructure investment that frontier AI development requires. OpenAI's infrastructure spend far exceeds what any enterprise could justify independently. The gap suggests that most organizations will consume AI through APIs rather than building comparable infrastructure themselves. The scale advantages of concentrated AI infrastructure favor the hyperscalers and frontier labs with the resources to build at gigawatt scale.

Stargate also signals the power requirements that AI creates for the broader economy. Eight gigawatts represents a meaningful fraction of national electricity generation. As AI deployment accelerates, power consumption follows. Grid operators, utilities, and energy producers will see AI as an increasingly significant customer. The AI infrastructure buildout drives energy infrastructure investment as a downstream consequence.

References

1. CNN. "Stargate: Trump announces a $500 billion AI infrastructure investment in the US." January 21, 2025. https://www.cnn.com/2025/01/21/tech/openai-oracle-softbank-trump-ai-investment

2. OpenAI. "OpenAI, Oracle, and SoftBank expand Stargate with five new AI data center sites." September 23, 2025. https://openai.com/index/five-new-stargate-sites/

3. Intuition Labs. "OpenAI's Stargate Project: A Guide to the AI Infrastructure." 2025. https://intuitionlabs.ai/articles/openai-stargate-datacenter-details

4. Intuition Labs. "OpenAI's Stargate Project."

5. Wikipedia. "Stargate LLC." https://en.wikipedia.org/wiki/Stargate_LLC

6. Wikipedia. "Stargate LLC."

7. Wikipedia. "Stargate LLC."

8. Intuition Labs. "OpenAI's Stargate Project."

9. Intuition Labs. "OpenAI's Stargate Project."

10. CNBC. "OpenAI's first data center in $500 billion Stargate project is open in Texas." September 23, 2025. https://www.cnbc.com/2025/09/23/openai-first-data-center-in-500-billion-stargate-project-up-in-texas.html

11. Data Center Frontier. "Scaling Stargate: OpenAI's Five New U.S. Data Centers Push Toward 10 GW AI Infrastructure." September 2025. https://www.datacenterfrontier.com/machine-learning/article/55319132/scaling-stargate-openais-five-new-us-data-centers-push-toward-10-gw-ai-infrastructure

12. OpenAI. "OpenAI, Oracle, and SoftBank expand Stargate with five new AI data center sites."

13. Data Center Frontier. "Scaling Stargate."

14. Data Center Frontier. "Scaling Stargate."

15. Intuition Labs. "OpenAI's Stargate Project."

16. Intuition Labs. "OpenAI's Stargate Project."

17. Certrec. "Energy Demands for OpenAI's Stargate Project." 2025. https://www.certrec.com/blog/energy-demands-for-openai-stargate-project/

18. Certrec. "Energy Demands for OpenAI's Stargate Project."

19. OpenAI. "Expanding Stargate to Michigan." 2025. https://openai.com/index/expanding-stargate-to-michigan/

20. Wikipedia. "Stargate LLC."

21. Wikipedia. "Stargate LLC."

22. TechTarget. "Stargate AI explained: What's in the $500 billion project." 2025. https://www.techtarget.com/whatis/feature/Stargate-AI-explained-Whats-in-the-project

Key takeaways

For strategic planners: - $500B over 4 years; 8GW capacity and $450B committed by late 2025, ahead of original schedule - SoftBank sold all NVIDIA shares ($30B) to fund Stargate; AMD providing GPUs for 6GW future capacity - International expansion: Stargate UAE (1GW Abu Dhabi), Norway (hydro-powered), UK, Argentina ($25B Patagonia)

For infrastructure architects: - Abilene flagship: 1.2GW across 8 AI factory halls; 400,000 GB200 GPUs planned; 16,000 installed by mid-2025 - Rack density 100-150kW; direct-to-chip liquid cooling within closed-loop systems mandatory - Five additional sites: Shackelford County TX, Doña Ana County NM, Lordstown OH, Milam County TX, Wisconsin

For finance teams: - Ownership: SoftBank 40% ($19B), OpenAI 40% ($19B), Oracle ($7B), MGX ($7B) - NVIDIA signed $100B investment to purchase AI processors for the data centers - 8GW represents meaningful fraction of national electricity generation; grid interconnection timelines extend years

For operations teams: - Power strategy: SMR nuclear for baseload, on-site solar, 360MW natural gas turbines backup - Operates on Oracle Cloud Infrastructure; Abilene receiving GB200 racks since mid-2025 - Grid strain: 8GW new load requires generation, transmission, and distribution capacity that may not exist

SEO Elements

Squarespace Excerpt (159 characters): OpenAI's Stargate reaches 8 gigawatts and $450 billion committed. 400,000 GPUs at Abilene alone. Inside the largest AI infrastructure project in history.

SEO Title (56 characters): OpenAI Stargate: Inside the $500 Billion AI Infrastructure

SEO Description (155 characters): OpenAI's Stargate project reaches 8 gigawatts of AI capacity across Texas, Ohio, and New Mexico. Analysis of the $500 billion infrastructure scale and power requirements.

URL Slugs: - Primary: openai-stargate-500-billion-ai-infrastructure - Alt 1: stargate-project-openai-softbank-oracle-2025 - Alt 2: stargate-data-center-10-gigawatt-capacity - Alt 3: trump-stargate-ai-infrastructure-announcement