Microsoft Expands Deployment of In-House Maia Accelerators to Lessen Nvidia Dependence

Original summary · What happened

Microsoft announced on Monday that it will expand the deployment of its proprietary artificial intelligence processors, moving to lessen its operational reliance on graphics processing units supplied by Nvidia. The technology conglomerate introduced its third-generation Maia AI accelerator, stating that the proprietary silicon will serve as the primary hardware running AI inference tasks to service incoming customer requests across its broader product ecosystem.

While Nvidia's hardware remains the prevailing industry standard across major data centers and model training workflows for developers like OpenAI and Anthropic, major hyperscalers have progressively diversified their silicon footprints. Microsoft currently supports cloud services and productivity assistants such as Microsoft 365 Copilot and GitHub Copilot via Azure. Alongside third-party processors sourced from Advanced Micro Devices, Intel, Amazon, and Google, expanding the internal Maia platform is intended to address soaring computational demand and contain infrastructure expenditures.