Project Stargate: OpenAI, SoftBank, Oracle & MGX Mobilize $500B Supercomputing Grid for Frontier AI
WASHINGTON, D.C. & SAN FRANCISCO — August 22, 2026 — In what stands as the largest private infrastructure capital deployment in human history, OpenAI, SoftBank Group, Oracle, and investment powerhouse MGX have formally established Stargate LLC, a $500 billion joint venture committed to engineering, financing, and constructing a nationwide 10-gigawatt AI supercomputing grid.
1. The Scale of Stargate: An Industrial-Age Manhattan Project for AI
As frontier artificial intelligence models demand exponentially higher compute for post-training reinforcement learning, autonomous agentic loops, and test-time reasoning, the global computational bottleneck has shifted from raw silicon supply to land, power generation, and specialized high-voltage grid infrastructure.
Project Stargate addresses this challenge at a planetary scale. Chaired by Masayoshi Son of SoftBank, the initiative unites SoftBank's capital orchestration, Oracle's high-density bare-metal RDMA networking, OpenAI's model architectures, and MGX's sovereign co-investment, supported by technology partnerships with NVIDIA, Microsoft, and Arm.
"This is the Apollo program for the intelligence age. With Stargate, we are creating the physical and computational foundation required to train models that will solve humanity's deepest challenges across energy, materials science, and medicine."
2. Flagship Texas Groundbreaking & Multi-State Expansion
The flagship Stargate campus has broken ground in Abilene, Texas, with initial power allocations exceeding 1.2 gigawatts, supported by on-site dedicated natural gas turbines, battery storage reserves, and direct grid interconnects with ERCOT.
Following the Texas deployment, Stargate LLC announced simultaneous multi-gigawatt facilities breaking ground across Ohio, New Mexico, and Wisconsin, targeting a total cumulative capacity of 10 gigawatts by 2029.
Core Pillars of the Stargate LLC Joint Venture
3. Powering the Era of Autonomous Reasoning Agents
Sam Altman, CEO of OpenAI, emphasized that the transition from static LLM chatbots to autonomous, long-horizon reasoning agents requires continuous, massive test-time compute. Supercomputing fabrics of this magnitude allow millions of agentic threads to verify hypotheses, execute sandboxed code, and synthesize new scientific discoveries in real-time.
By securing domestic, high-density energy pipelines and next-generation liquid cooling systems, Project Stargate establishes the technological sovereignty and computational capacity required to lead the world into the era of artificial general intelligence.