传OpenAI考虑租赁英伟达(NVDA.US)芯片，或节省10%-15%支出

Core Insights - OpenAI is exploring the feasibility of leasing AI chips from NVIDIA to reduce cost pressures in their large-scale data center collaboration [1] - By leasing server chips instead of purchasing them outright, OpenAI could save 10% to 15% on costs [1] - The new leasing agreement with NVIDIA may have a five-year term, similar to existing agreements with Oracle [1] Group 1: OpenAI and NVIDIA Collaboration - OpenAI has been leasing NVIDIA chips from cloud service providers like Microsoft and Oracle [1] - NVIDIA plans to invest up to $100 billion in supporting OpenAI to build and deploy AI data centers with a capacity of at least 10 gigawatts [1] - The first phase of this project is expected to start in the second half of 2026, utilizing NVIDIA's Vera Rubin platform [1] Group 2: Future Plans and Infrastructure - Altman emphasized the potential of having 10 gigawatts of computing power to find cures for diseases like cancer and provide personalized education globally [1] - There is a vision to establish a capacity of 1 gigawatt of AI infrastructure added weekly to continuously expand computing power [2] - More details on future plans and partnerships will be disclosed in the coming months, including specific financing strategies [2]