Microsoft Finally Announces Proprietary Chips Focused on AI

At the recent Microsoft Ignite event, the tech company lifted the veil on a pair of newly designed chips and integrated systems. The company announced the Microsoft Azure Maia AI Accelerator, fashioned specifically for AI applications and generative AI tasks, and the Microsoft Azure Cobalt CPU, an Arm-based processor engineered to execute general-purpose compute operations on the Microsoft Cloud.

These chips serve as the final missing link, enabling Microsoft to offer wholly integrated infrastructure systems. These systems, which involve a range of components — from “silicon choices, software and servers to racks and cooling systems” — have been designed from the ground up for optimization with both internal and client workloads.

“Microsoft is building the infrastructure to support AI innovation, and we are reimagining every aspect of our datacenters to meet the needs of our customers. At the scale we operate, it’s important for us to optimize and integrate every layer of the infrastructure stack to maximize performance, diversify our supply chain and give customers infrastructure choice.”

Scott Guthrie, executive vice president of Microsoft’s Cloud + AI Group

Microsoft plans to deploy these chips in its data centers as early as next year. Initially, they will power services such as Microsoft Copilot and the Azure OpenAI Service. These chips will supplement a growing array of products from industry partners, all aimed at satisfying the burgeoning need for efficient, sustainable, and scalable computation capabilities. This move is a boon for clients keen on leveraging the latest advancements in cloud and AI technology.

During the Microsoft Ignite event, the tech mammoth announced the public availability of Azure Boost, a unique system designed to enhance networking and storage speeds. It achieves this by relieving the host servers of these processes and assigning them to specially designed hardware and software.

In a bid to bolster its custom silicon initiatives, Microsoft also revealed an expansion of its industry partnerships. This strategic move is aimed at providing customers with a broader range of infrastructure alternatives. One significant revelation was the launch of a trial version of the NC H100 v5 Virtual Machine Series. This product has been tailored for NVIDIA H100 Tensor Core GPUs and promises improved performance, reliability, and efficiency for mid-range AI training and generative AI inferencing. Furthermore, Microsoft announced plans to integrate the latest NVIDIA H200 Tensor Core GPU into its lineup next year, which will facilitate larger model inferencing without latency escalation.

In another strategic move, Microsoft revealed the upcoming addition of AMD MI300X accelerated VMs to its Azure lineup. The ND MI300 virtual machines are specifically designed to expedite AI workloads for high-range AI model training and generative inferencing. These VMs will be powered by AMD’s newest GPU, the AMD Instinct MI300X.

Going forward, “Microsoft plans to expand that set of options in the future; it is already designing second-generation versions of the Azure Maia AI Accelerator series and the Azure Cobalt CPU series.”