Group 1 - OpenAI plans to deploy one million GPUs by the end of this year, significantly increasing its GPU capacity from the current 100,000 GPUs used for ChatGPT 4.5 [1][2] - The surge in GPU deployment indicates a corresponding increase in demand for AI servers, with major suppliers like Hon Hai, Quanta, and Inventec expected to benefit from this trend [1][2] - The "Stargate" initiative by the Trump administration is anticipated to further drive AI infrastructure demand, leading to a substantial influx of AI server orders [1] Group 2 - OpenAI's transition to a "multi-cluster architecture" is aimed at supporting the training needs of next-generation models, with future considerations for using up to 10 million GPUs simultaneously [2] - The Stargate project, with a budget of hundreds of billions, is progressing, with the first data center in Texas expected to house up to 400,000 NVIDIA AI chips, potentially becoming a leading AI computing center globally [2] - OpenAI's cloud service providers include Microsoft, Oracle, and CoreWeave, with major Taiwanese suppliers like Hon Hai and Quanta set to gain from increased server production [3]
OpenAI冲刺算力 规模将增长十倍 鸿海、广达等将迎大单
Jing Ji Ri Bao·2025-07-21 23:21