OpenAI首席执行官Sam Altman正在展开一场全球性的资金筹集和供应链布局行动，寻求融资和制造合作伙伴以满足这家初创公司对算力的巨大需求。 与2024 年初那次备受质疑的7万亿美元基础设施计划推介不同，这一次他获得了更多支持。 10月4日，据报道，自9月下旬以来，这位ChatGPT制造商的掌门人已先后到访中国台湾、韩国和日本，以加速全球人工智能芯片制造能力建设。 他与台积 电、富士康、三星和SK海力士等公司会面，推动这些公司增加产能并优先处理OpenAI的订单。 Altman还计划访问阿联酋的投资者，为OpenAI的基础设施扩张和研究筹集资金。据知情人士透露，潜在的新资本将部分用于资助位于阿布扎比的Stargate（星 际之门）数据中心。 这次行程的背景是OpenAI与英伟达达成的重磅协议，芯片巨头同意向这家ChatGPT制造商租赁多达500万颗AI芯片，并投入多达1000亿美元来实现这一目 标。该协议提升了全球芯片供应商的股价，并为Altman的"算力帝国"愿景提供了有力支撑。 从质疑到支持：算力需求获得认可 Altman此次出访让人想起他在2024年初的一次行程。当时，他向这些公司（台积电、三星、 ...

Core Insights - OpenAI CEO Sam Altman is actively seeking funding and manufacturing partners to meet the company's significant demand for computing power, with recent trips to Taiwan, South Korea, and Japan to enhance global AI chip manufacturing capabilities [1][2] - OpenAI's valuation has reached $500 billion, comparable to major global companies like Netflix and ExxonMobil, following a substantial agreement with Nvidia for up to 5 million AI chips and an investment of up to $100 billion [2][3] Group 1: Funding and Partnerships - Altman is engaging with major companies such as TSMC, Foxconn, Samsung, and SK Hynix to increase production capacity and prioritize OpenAI's orders [1][3] - In the Middle East, Altman plans to meet with investors in Abu Dhabi to raise funds for OpenAI's infrastructure expansion, including the Stargate data center [4] Group 2: Chip Demand and Production Goals - OpenAI's overall demand for chips may reach 900,000 wafers per month, more than double the current global high-bandwidth memory capacity [3] - The company aims to create a factory capable of producing 1 GW of new AI infrastructure weekly, with plans to deploy at least 10 GW of Nvidia computing systems for training and running next-generation models [6] Group 3: Strategic Collaborations - OpenAI has established partnerships with South Korean companies for storage chip collaboration and with Hitachi in Japan to support AI infrastructure development [3] - Discussions are ongoing regarding the manufacturing and deployment of Nvidia's upcoming Rubin system, with OpenAI set to be one of the first customers [3]