Core Viewpoint - Microsoft has officially launched its first "AI Super Factory," connecting data centers across states to create a distributed computing cluster, marking a shift towards a networked collaborative era in AI infrastructure [1][3]. Group 1: AI Super Factory Concept - The "AI Super Factory" integrates geographically dispersed data centers into a virtual supercomputer, differing from traditional data center designs [5]. - This distributed network connects multiple sites, consolidating advanced GPUs, EB-level storage, and millions of CPU cores to support future AI model training with trillions of parameters [5][6]. Group 2: Infrastructure and Technology - The new data center in Atlanta spans 85 acres with over 1 million square feet designed specifically for AI workloads [6]. - Key features include a high-density architecture, advanced chip systems with NVIDIA GPUs, an efficient liquid cooling system, and high-speed internal connectivity [7][8]. Group 3: Network and Power Distribution - Microsoft has deployed 120,000 miles of dedicated fiber optic cables to create a high-speed network for AI traffic, ensuring data transmission at near-light speed [9]. - The decision to build across states is driven by land and power supply considerations, allowing for distributed power demand and avoiding overloading any single grid [9]. Group 4: Competitive Landscape - Microsoft’s "AI Super Factory" is a strategic asset in the escalating competition for AI computing power, with plans to double its data center footprint in the next two years to meet surging demand [4][10]. - Competitors like Amazon, Meta Platforms, and Oracle are also ramping up their infrastructure investments, indicating a broader industry trend towards massive AI-related capital expenditures [10].
微软第一座“AI超级工厂”投入运营:将两座数据中心连接,构建分布式网络