BroadChain has learned that on March 17, OpenAI significantly revised its Stargate computing infrastructure strategy, pivoting away from building its own data centers. Instead, the company will now rely on third-party cloud services, including AWS and Google Cloud, to meet its computational needs.
As part of this strategic shift, OpenAI has also lowered its projected total computing expenditure through 2030 from an initial $1.4 trillion to approximately $600 billion.
The revision is largely attributed to financing challenges. Since its initial proposal in 2025, the Stargate project has faced slow progress and struggled to secure the large-scale funding required for construction.
OpenAI has withdrawn from related expansion negotiations in Texas and has established a new collaboration framework with SoftBank. Under this agreement, SoftBank will manage energy and facility development, while OpenAI will secure computing resources through long-term leasing contracts.
Furthermore, OpenAI is reorienting its computational focus toward NVIDIA’s Vera Rubin platform, with the goal of achieving its first gigawatt-scale computing capacity in the second half of 2026.