Microsoft to Prioritize In-House AI Chips for Data Centers

Microsoft Corporation plans to primarily use its own custom-designed chips in its data centers, as confirmed by Chief Technology Officer Kevin Scott. This strategic shift aims to reduce reliance on major chip manufacturers like Nvidia and AMD.

Currently, Microsoft utilizes a mix of Nvidia, AMD, and its own custom silicon in its data centers. The company has been focusing on selecting chips that offer the best price performance. However, the explosive growth in AI demand has prompted Microsoft to develop its own chips, such as the Azure Maia AI Accelerator and the Cobalt CPU, to better meet its specific requirements.

Scott emphasized that the focus is not just on individual chips but on designing entire data center systems, including networking and cooling solutions. This approach aims to optimize computing power according to workload needs. Despite significant investments in AI infrastructure, Microsoft faces a substantial computing power gap, with demand outpacing supply since the launch of AI models like ChatGPT.

The move to in-house chip development is part of a broader industry trend among tech giants like Google and Amazon, who are also designing their own chips to enhance efficiency and reduce dependency on external suppliers.