New Projects Add Over 4.5 GW to Capacity Pipeline
OpenAI is accelerating its global data center strategy with a newly announced 4.5 gigawatt expansion in collaboration with Oracle, adding to an 800 MW facility already underway in Abilene, Texas. The initiative, operating under the “Stargate” brand, reflects OpenAI’s growing need for computing power to support its next-generation AI models, such as GPT-5. Despite setbacks with SoftBank and delays in other projects, OpenAI remains committed to aggressive infrastructure development.
In total, the Stargate initiative now includes four publicly known sites: the original Abilene facility, the new 4.5 GW U.S. expansion (currently in site assessment), a 1 GW facility in the UAE with multiple global partners, and an unannounced 250 MW site in Denton, Texas. These projects involve partners like Oracle, Nvidia, Cisco, SoftBank, G42, and Core Scientific.
Supply Chain and Partnership Challenges Emerge
OpenAI’s ambitions to scale up infrastructure face significant headwinds. The company has yet to finalize concrete deals under its $100 billion pledge with SoftBank, with only a small U.S. data center likely to move forward in the near term. SoftBank, despite trademarking “Stargate,” is absent from the latest Oracle-led U.S. expansion, signaling strained relations and project realignments. Meanwhile, logistical constraints and strained supply chains are slowing the pace of development, according to internal memos shared with employees.
Sam Altman acknowledged that OpenAI’s demand for compute is “starting to strain the supply chain” and will “require some real creativity.” The company expects to have 2 million Nvidia A100 GPUs available by the end of August and aims to double that number by year’s end, signaling a continued push for scale despite infrastructure bottlenecks.
Global Expansion via Strategic Alliances
OpenAI is diversifying its compute supply through partnerships beyond its traditional relationship with Microsoft. It has signed its first international agreement through the “OpenAI for Countries” program with the United Arab Emirates and is pursuing additional deals in Asia. Its Middle East project, Stargate UAE, features collaborations with Oracle, Nvidia, Cisco, and AI startup G42, backed by Microsoft.
In parallel, OpenAI is reportedly seeking compute resources from rival Google and continuing talks with international governments and tech companies to expand its infrastructure. Jason Kwon, OpenAI’s Chief Strategy Officer, has been leading outreach efforts in countries like India, Korea, and Singapore.
Industry-Wide Scramble for Compute Dominance
The race for AI supremacy is fueling an unprecedented surge in global data center demand. Morgan Stanley projects a 23% annual increase in data center capacity through 2030, driven by leading players including Meta, Google, Microsoft, and Elon Musk’s xAI. Musk recently revealed that xAI has over 230,000 GPUs in operation, with 550,000 Nvidia chips expected online soon for training its Grok model.
Anthropic, another major AI contender, is reportedly seeking funding from Middle Eastern governments to finance its data center needs, despite ethical concerns. Co-founder Dario Amodei noted in a staff memo that idealism may need to yield to practicality in order to scale effectively.