By Arnab Paul
Publication Date: 2026-10-12 17:35:00
Microsoft expects its next-generation Maia 300 system to deliver 40% more AI-generated tokens per megawatt than ‘leading GPU-based systems.’
- The company will begin deploying Maia 300 chips from 2027.
- Maia 200 already powers Microsoft Copilot with GitHub Copilot, and Microsoft Foundry is expected to follow by year-end.
- Maia 200 is currently being rolled out across the U.S., Europe, and Australia.
Microsoft (MSFT) is ramping up its push into homegrown AI chips, unveiling plans to deploy its Azure Maia systems at scale from next year, as the technology giant looks to cut costs and reduce its dependence on external chipmakers amid surging AI demand.
On Monday, the company said it will scale Maia to multi-gigawatt deployments beginning in 2027, making it the primary system for running AI models across Microsoft products and services. Maia is Microsoft’s AI chip developed to run large AI models in its Azure cloud data centers.
MSFT shares were up over 1% at the time of writing. The…