英伟达与OpenAI达成千亿美元投资，加码AI算力基建

Group 1 - Nvidia announced plans to invest up to $100 billion in OpenAI to build AI data centers with a power capacity of 10 gigawatts [2] - The construction of the 10-gigawatt data center will require the deployment of 4 to 5 million GPUs, equivalent to Nvidia's total shipments this year or double last year's shipments [2] - OpenAI and Nvidia's partnership dates back to 2016, when Nvidia delivered its first AI supercomputer to OpenAI [2] Group 2 - The collaboration between Nvidia and OpenAI is part of the "Stargate" project, which involves a $500 billion investment in AI infrastructure over four years by Oracle, OpenAI, and SoftBank [3] - OpenAI has recently increased the power capacity of the Stargate project by an additional 4.5 gigawatts through a separate agreement with Oracle [3] - Reports indicate that OpenAI plans to purchase $300 billion worth of computing power over the next five years, raising questions about the feasibility given OpenAI's annual revenue of approximately $10 billion [3] Group 3 - The relationship among Nvidia, OpenAI, and Oracle is characterized as a closed-loop system where Nvidia provides funding, OpenAI purchases cloud computing services from Oracle, and Oracle procures chips from Nvidia [3] - Since Microsoft's investment in OpenAI in 2019, OpenAI has primarily used Microsoft's Azure cloud services, but its deepening collaboration with Oracle is seen as a move to reduce dependency on Microsoft [4] - OpenAI is also exploring the development of its own chips to lessen its reliance on Nvidia [4]