Microsoft has built its first custom silicon designs, announcing two chips named Azure Maia 100 and Cobalt 100. Both chips are described as purpose built for Artificial Intelligence and cloud workloads and are part of Microsoft’s effort to deploy in-house processors inside its Azure infrastructure. The company positions these as the first step in a broader move to tailor hardware to the demands of its cloud services.
The Azure Maia 100 is presented as a graphics processing unit that will assist in Artificial Intelligence tasks, while the Cobalt 100 is a central processing unit aimed at general cloud services. The article specifies that Cobalt 100 contains 128 Arm-based cores, highlighting its focus on large-scale, general compute within Microsoft’s data centers. Microsoft says the chips are intended to optimize performance in Azure, with availability planned for 2024.
Microsoft frames the designs as a way to improve its cloud infrastructure and reduce reliance on existing third-party accelerators. The company expects the custom silicon to enhance the efficiency and capability of Azure data centers when running Artificial Intelligence and other cloud workloads. The announcement is described as part of Microsoft’s continued evolution in cloud and Artificial Intelligence capabilities, signaling a strategic push to integrate more in-house hardware into its platform.
