Microsoft's Maia 300 rack: 96 accelerators, HBM4, mostly liquid-cooled
Microsoft has released architecture details for Maia 300, its third-generation Azure AI accelerator. The design pairs 96 accelerators per rack with a two-tier network that links up to 12,288 accelerators into a single superpod. The company says the system is built for cloud-scale agentic inference, and that it co-designed the software, silicon, rack infrastructure, networking, and data centers to run as one platform.
Read original article ↗
Related Articles
Microsoft to scale Maia AI chips to multi-gigawatt deployments, unveils TSMC N2P-based Maia 300
Microsoft will scale its in-house Azure Maia AI accelerators to multi-gigawatt deployments starting in 2027, making Maia
Microsoft’s new AI chip makes it less reliant on Nvidia’s GPUs
Microsoft’s Maia 300 chip. | Image: Microsoft Microsoft is announcing its third-generation Maia in-house AI chip today
Anugal Brings Agentic AI Into Enterprise Identity Governance and Access Decisions
From AI-assisted SAP role design to intelligent access reviews and Microsoft Teams, Anugal helps enterprises make faster