While OpenAI races to build AI data centers, Nadella reminds us that Microsoft already has them (techcrunch.com)

Microsoft CEO Satya Nadella tweeted a video debuting the company’s first deployed massive AI “factory” — a cluster of more than 4,600 Nvidia GB300 rack servers built around Nvidia’s new Blackwell Ultra GPUs and linked with Nvidia’s InfiniBand high‑speed networking. Microsoft says this is the “first of many” such systems to be rolled out across Azure, with plans to deploy “hundreds of thousands of Blackwell Ultra GPUs” globally. The systems sit inside Microsoft’s existing footprint of 300+ data centers in 34 countries and are advertised as capable of running next‑generation models with “hundreds of trillions of parameters.” The announcement is significant because it signals Microsoft’s strategy to meet frontier AI compute demand at scale while positioning itself against OpenAI, which is concurrently signing major data‑center deals with Nvidia and AMD and pledging its own large buildouts. Technical takeaways include Microsoft leveraging Nvidia’s end‑to‑end stack (chips plus InfiniBand, the latter strengthened by Nvidia’s 2019 Mellanox acquisition) to optimize throughput and model parallelism for enormous models. The move underscores that cloud providers—not only AI labs—are rapidly building vertically integrated GPU farms to host and monetize large models, and it sets expectations for forthcoming details on Azure’s AI offerings at upcoming industry events.