Microsoft has officially launched its first large-scale artificial intelligence system, characterized as an "AI factory," within its global Azure data center infrastructure. Showcased by CEO Satya Nadella, the deployment is intended to support OpenAI workloads and represents the initial phase of what the company described as "the first of many" Nvidia-powered AI factories slated for global expansion.
Each of these new AI systems is configured as a cluster containing over 4,600 Nvidia GB300s rack computers. These units are equipped with Nvidia's Blackwell Ultra GPU chips and leverage Nvidia's InfiniBand networking technology for high-speed connectivity. Microsoft stated its plans include deploying "hundreds of thousands of Blackwell Ultra GPUs" as it continues to roll out these advanced systems across its worldwide network. Nvidia's acquisition of Mellanox in 2019 for $6.9 billion secured its position in the InfiniBand market.
This announcement follows recent high-profile data center agreements by OpenAI, Microsoft's partner, with both Nvidia and AMD. Reports indicate OpenAI has accumulated approximately $1 trillion in commitments to develop its own data centers in 2025, with CEO Sam Altman confirming additional deals are anticipated.
Microsoft emphasized its existing global data center footprint, which encompasses over 300 facilities across 34 countries. The company asserted these data centers are "uniquely positioned" to "meet the demands of frontier AI today." The newly deployed AI systems are engineered to handle the next generation of models, capable of managing "hundreds of trillions of parameters," according to company statements.
Further details on Microsoft's strategy for scaling AI workloads are expected later this month. Microsoft CTO Kevin Scott is scheduled to speak at TechCrunch Disrupt, taking place from October 27 to October 29 in San Francisco.
