OpenAI要买40万个GPU

Core Insights - OpenAI is investing $100 billion in the Stargate infrastructure project, which will feature a data center capable of housing up to 400,000 Nvidia AI chips, potentially making it one of the largest AI computing clusters known [1] - The data center, located in Abilene, Texas, is expected to be completed by mid-2026 and will have a power capacity of 1.2 gigawatts [1] - The Stargate joint venture was announced in January and aims to provide the physical infrastructure needed for advanced AI models [1] Project Development - Oracle has agreed to dedicate the entire Abilene facility to the Stargate project, with OpenAI planning to utilize approximately 1 gigawatt of capacity [2] - The project has already begun construction, with about 2,000 workers currently involved and plans to increase to nearly 5,000 [3] - The facility will consist of eight data center buildings, each designed to accommodate up to 50,000 Nvidia GB200 semiconductors [3] Competitive Landscape - The Stargate project is part of a broader competition among leading tech companies to expand the production capacity of Nvidia's latest chips [2] - Other companies, such as Elon Musk's xAI and Meta Platforms Inc., are also making significant investments in AI infrastructure, with xAI signing a $5 billion deal for AI server provision and Meta planning to achieve computing capacity equivalent to 600,000 Nvidia H100 chips by the end of 2024 [2]