Microsoft CEO Satya Nadella tweeted a video on Thursday showcasing the company's first deployment of a large-scale AI system—what Nvidia likes to call an AI "factory." He promised that this was the "first" of "many" Nvidia AI factories to be deployed in Microsoft Azure datacenters worldwide, running OpenAI workloads.

Each system is a cluster of over 4,600 Nvidia GB300 rack computers, powered by its highly sought-after Blackwell Ultra GPU chips, connected by Nvidia's ultra-high-speed networking technology called InfiniBand. (Besides AI chips, Nvidia CEO Jensen Huang also had the foresight to corner the InfiniBand market with his $6.9 billion acquisition of Mellanox in 2019.)

Microsoft promises that as these systems roll out globally, they will deploy "hundreds of thousands of Blackwell Ultra GPUs." While the scale of these systems is impressive (and the company has shared more technical details for hardware enthusiasts to explore), the timing of the announcement is also noteworthy.

The news comes just after OpenAI, Microsoft's partner and notorious frenemy, signed two high-profile data center deals with Nvidia and AMD. It's estimated that OpenAI has secured $1 trillion in commitments to build its own data centers through 2025. CEO Sam Altman said this week that more deals are on the horizon.

Microsoft clearly wants to make it clear that it has the data centers—over 300 in 34 countries—and is "uniquely positioned" to "meet the demands of today's cutting-edge AI." The company says these massive AI systems are also capable of running next-generation models with "hundreds of trillions of parameters."

We expect to hear more about how Microsoft is accelerating AI workloads later this month. Microsoft CTO Kevin Scott will speak at TechCrunch Disrupt, the San Francisco conference from October 27th to 29th.

Q&A

Q1: What is the specific configuration of the AI ​​factory Microsoft is deploying?

A: Each AI Factory system is a cluster of over 4,600 NVIDIA GB300 rack computers, powered by Blackwell Ultra GPU chips, connected via NVIDIA's ultra-high-speed InfiniBand networking technology. Microsoft has pledged to deploy hundreds of thousands of Blackwell Ultra GPUs.

Q2: Why is Microsoft announcing the AI ​​data center construction now?

A: This announcement comes just after OpenAI signed large-scale data center agreements with NVIDIA and AMD, securing a $1 trillion commitment to build its own data centers. Microsoft wants to highlight its already dominant position with over 300 data centers in 34 countries.

Q3: What scale of models can these AI systems handle?

A: Microsoft claims these massive AI systems are capable of running next-generation models with "hundreds of trillions of parameters" and are uniquely positioned to meet the demands of today's cutting-edge AI, effectively serving AI workloads like OpenAI's.