Microsoft Corp. is making significant moves to integrate its own chips into data centers, marking a major step toward powering its AI software with in-house hardware. The tech giant plans to deploy multiple gigawatts of its Maia chips across its server farms, with the latest version, the Maia 300, set to begin rolling out next year. These semiconductors are specifically designed to handle inference tasks, such as summarizing conference calls or generating answers to research questions.
The announcement came on Monday, underscoring Microsoft's push to rely more on its proprietary technology for AI applications. The Maia 300 chips represent the newest iteration in the company's efforts to reduce dependence on third-party hardware providers. This strategic shift could enhance performance and control over AI-driven services.