Microsoft launches two custom datacenter infrastructure chips to accelerate AI application performance.

Microsoft announced two additional infrastructure chips for its data centers at its Ignite conference on November 20, 2024, designed to speed artificial intelligence operations and increase data security. Like rivals Amazon.com and Google, Microsoft has devoted significant resources to develop home-grown silicon for general purpose applications and artificial intelligence, as engineers say there is a performance and price benefit to designing chips customized for its needs. Designing custom chips can reduce Microsoft's reliance on processors made by Intel and Nvidia.

The first new chip, called the Azure Integrated HSM, is designed to increase security by keeping crucial encryption and other security data inside the security module. Engineers will install this security chip in every new server destined for a data center beginning next year. The second chip is a data processing unit, or DPU, aimed at cloud storage data processing, which moves multiple components of a server into a single chip focused on that function.

According to Rani Borkar, corporate vice president, Azure Hardware Systems and Infrastructure, the company makes the effort to design an array of data center processors because it aims to "optimize every layer of infrastructure" and ensure that Microsoft's data centers crunch information at the speed AI requires. The DPU can run these specific tasks at three times less power and four times the performance compared with Microsoft's current hardware.

In addition to the chips, Microsoft announced a new version of a cooling system for data center servers that relies on liquid to reduce the temperature of nearby components. The cooling unit can be used to support large-scale AI systems, further contributing to Microsoft's infrastructure optimization efforts.