aiqbench takeChips & Infra

Microsoft's shift to deploying its proprietary Maia chips at a multi-gigawatt scale marks a significant move toward vertical integration for AI inference workloads. This strategy reduces reliance on external silicon suppliers and aims to optimize performance specifically for Microsoft's own large-scale AI models. The upcoming rollout of the Maia 300 generation suggests continued iteration in hardware efficiency tailored to specific inference tasks like summarization and research assistance.

Source covering36Kr

What this isA model's reading of the article and related coverage, written for aiqbench. Not the publisher's words.

Microsoft Prepares to Deploy Self-Developed Chips on a Gigawatt Scale in Data Centers

36KrOctober 12, 20261 min read

Microsoft is preparing to deploy its latest self-developed chips in data centers, aiming to rely primarily on its own hardware to provide computing power for its AI software. Microsoft stated on Monday that it plans to deploy Maia chips with a total capacity of several gigawatts in its data centers, with the newest generation, Maia 300, set to begin deployment next year. These chips are designed specifically for AI inference, enabling tasks such as summarizing conference calls or answering research questions using AI models.

Read the full article at 36Kr

Read at 36Kr ↗Ask about this article