Microsoft has launched its newest Azure AI data center in Atlanta, Georgia, described as the world’s first “planet-scale AI super factory.” The Atlanta facility, the second in the line of Fairwater AI data centers, joins the first Fairwater site in Wisconsin, and connects to Microsoft’s broader global network of Azure data centers, enabling unprecedented compute density to power next-generation artificial intelligence systems.
What is Fairwater?
Readers would recall that in September 2025, Microsoft announced the Wisconsin AI data center as the first in line of many subsequent “Fairwater” facilities that were under construction elsewhere across the country.
The Fairwater design is a departure from traditional cloud data centers since each site uses a single flat network capable of integrating hundreds of thousands of NVIDIA GB200 and GB300 Graphic Processing Units (GPUs) into a massive unified supercomputer. The data center architecture allows for efficient scaling of diverse AI workloads from pre-training and fine-tuning to reinforcement learning and synthetic data generation.
Fairwater’s liquid-cooling system uses a closed-loop design that continuously reuses coolant with minimal water consumption. This system enables high rack-level power density (140kW per rack) while maintaining efficiency and stability for large-scale compute sustainability.
The data center’s two-story architecture further reduces cable lengths and latency between GPUs, improving performance and reliability for sensitive AI workloads.
The clusters are linked through a two-tier Ethernet-based backend network using Microsoft’s SONiC operating system that offers 800 Gbps connectivity and scalable, vendor-agnostic performance. Microsoft’s AI WAN optical network extends Fairwater’s reach across 120,000 new miles of fiber laid in the United States of America (USA), which interlinks multiple data centers into one cohesive AI supercomputing ecosystem.
In a press release Microsoft stated that the company has deployed a dedicated AI WAN backbone, providing flexible workload allocation and maximizing GPU utilization across its global infrastructure to link Fairwater sites into a single cohesive system.
Satya Nadella, Chairman and CEO at Microsoft said in a Linkedin Post, “Fairwater exemplifies our vision for a fungible fleet: infra that can serve any workload, anywhere, on fit-for-purpose accelerators and network paths, with maximum performance and efficiency.”
Microsoft selected Atlanta for its highly available and cost-efficient grid power capable of achieving 99.99% uptime at a fraction of traditional redundancy costs. The company has co-developed power-management technologies to stabilize grid demand, using intelligent software, GPU-level regulation, and on-site energy storage.