TMTPOST --   Microsoft Corp. is   launching a new   class of datacenter   as the   software   giant   is   ramping up artifcial intelligence ( AI )   infrastructure   buildout.

Credit：Microsoft

Microsoft on Wednesday   announced its first AI superfactory   was brought   online   to accelerate AI breakthroughs and train new models on a scale that has previously been impossible.   The superfactory is   delivered as a set of two-story   structure in Atlanta, Georgia, the second   in Microsoft ’ s Fairwater   family. The new   purpose-built Fairwater   site of Azure AI   datacenter   is connected to our first Fairwater site in Wisconsin.  

The Fairwater-site   superfactory shares the same architecture and design as Microsoft ’ s investment in Wisconsin   in September.   The   later was labelled by   Microsoft as the   world ’ s most powerful   AI datacenter.   But these aren ’ t simply isolated buildings densely packed with sophisticated silicon and cooling techniques that use almost zero water.

These Fairwater AI datacenters are directly connected to each other – and eventually to others under construction throughout the U.S.   – with a new type of dedicated network allowing data to flow between them extremely quickly. This enables Fairwater sites located in different states to work together as an AI superfactory to train new generations of AI models far more quickly, accomplishing jobs in just weeks instead of several months.

Microsoft said   the network will connect multiple sites with hundreds of thousands of the most advanced graphic   processing units   ( GPUs )   running AI workloads, exabytes of storage and millions of CPU cores for operational compute tasks.

The new   Fairwater AI datacenters have a unique design that differentiates them from previous generations. It   features a new chip and rack architecture that delivers the highest throughput per rack of any cloud platform available today.   It boasts Nvidia ’ s   GB200 NVL72 rack-scale systems that can scale to hundreds of thousands of Nvidia   Blackwell GPUs. Its   two-story design   allows for greater GPU density   and intelligent networking   enables   fast communication among GPUs. It also features an   advanced liquid cooling that consumes almost zero water in its operations   and a new dedicated network linking it to AI compute clusters at other sites.

"This is about building a distributed network that can act as a virtual supercomputer for tackling the world ’ s biggest challenges in ways that you just could not do in a single facility," said Alistair Speirs, Microsoft general manager focusing on Azure infrastructure.

  "A traditional datacenter is designed to run millions of separate applications for multiple customers," Speirs   added. "The reason we call this an AI superfactory   is it ’ s   running one complex job across millions of pieces of hardware.   And   it ’ s   not just a single site training an AI model, it ’ s   a network of sites supporting that one job."

Microsoft didn ’ t specify   the cost   of   the superfactory   in   Fairwater   site, but the company   had disclosed   aggressive   AI spending in late October   when   it   released financial   results   for its   first   fiscal   quarter ended September 30.  

Microsot posted a steeper climb in spending than Wall Street anticipated, fueling anxieties about the high costs of providing AI infrastructure. The September quarter saw captial expenditure, or Capex, including finance leases, an indication of data center spending,   refreshed   a record at $34.9 billion,   while analysts projected to be $30.06 billion. The Capex   was   up about $10   billion,   or   more than 60%,   from the previous record set during the previous quarter,   and surged 74.5%   from a year ago.

Microsoft CEO Satya Nadella   said   his company will continue to increase our investments in AI across   both   capital and talent   as its planet-scale cloud and AI factory, together with Copilots across high value domains, is driving broad diffusion and real-world impact.

Hood said   at an earnings call that the Capex will increase again in the current fiscal quarter. She told analysts Microsoft still failed to address   a computing capacity   crunch   despite massive spending on AI datacenters. Demand for Azure services is "significantly ahead of the capacity we have available," she said. "I thought we were going to catch up," Hood added. "We are not."