SoftBank and OpenAI have reportedly scaled down their ambitious AI infrastructure plans under the Stargate Project, transitioning from a $500 billion vision to a smaller data center launch by the end of 2025. As per a Wall Street Journal report cited by Reuters, the strategic pivot comes in response to global economic constraints, rising construction costs, and mounting funding pressures. The new plan now involves a smaller-scale but operable AI-ready data center to keep up with the accelerated demand for GPU cloud capabilities. It also illustrates changed priorities as companies rethink major investments into AI, fully recognizing the current infrastructure bottleneck and energy constraints.
Stargate Project Faces Major Budget Revisions
Originally framed as a transformative AI venture, the Stargate Project was meant to deliver massive computational infrastructure to support AI scalability across sectors. However, SoftBank and OpenAI are now recalibrating their strategy due to increasing financial headwinds.
This move highlights the budgetary caution that’s now creeping into major AI investments. Although Softbank pledged, $3 billion a year for AI funding, the current market conditions have given rise to a need for both companies to focus, on speed, efficiency, and speed, in addition to simple scale. The continuing partnership is still holding strong, yet both companies are now acting from a course of expensive max speed in line with what the market and power grid can sustain.
GPU Cloud Demand Highlights Infrastructure Strain
One of the core reasons behind the scaled-back Stargate Project is the intense demand for GPU cloud resources. AI model training and deployment require massive GPU clusters, but existing infrastructure has struggled to keep up. According to McKinsey’s 2024 findings, colocation providers and GPU cloud firms are rapidly expanding, yet power supply remains a limiting factor.
Utility providers typically offer power in tranches of 15-25 megawatts, which significantly delays large-scale deployment. This has affected data center growth in critical regions such as Northern Virginia, where vacancy rates are near zero. As OpenAI and SoftBank look to stay competitive, their recalibrated plan seeks to balance current GPU cloud requirements with feasible infrastructure growth.
Data Center Delays Tied to Power and Supply Gaps
While the data center segment grows more important in the AI race, logistic obstacles are just as relentless. When you face limited suitable land to build on, limited access to high-voltage power, and long construction timelines, you are essentially stuck before you can even get ramped up. The change in the size of the Stargate Project reflects this reality.
An example is Northern Virginia, where the AI-driven data center activity is still at a torrid pace with nearly zero vacancies, indicating the speed at which demand is now exceeding available supply. In this climate, SoftBank and OpenAI’s decision to build a more manageable facility appears practical, timed to saleable reality instead of a speculative fantasy.
AI Sovereignty in Focus Despite Infrastructure Revisions
Despite trimming its ambitions, the Stargate Project remains a key part of the broader push for AI sovereignty. SoftBank brings financial muscle, while OpenAI contributes technological expertise. Announced in February 2025, their collaboration still aims to define strategic AI capabilities, even if through smaller milestones.
But the chance is growing that if infrastructure challenges are unchanged, big capital may decide to redirect into decentralized AI solutions that are unrelated to traditional data center models. It might be more resilient and scalable in locations with little land and centralized power. For now, both companies are continuing to work through the complicated maze of GPU cloud access, data center logistics, and the growing global AI competition, but at a slower cadence.
About The Author
