Microsoft Maia 300
Microsoft's third-generation in-house accelerator on TSMC N2P: eight HBM4 stacks, 288 GB, 18 TB/s per chip.
Microsoft’s third-generation in-house accelerator on TSMC N2P: eight HBM4 stacks, 288 GB, 18 TB/s per chip; a 167 kW rack of 96 accelerators gives up to 7x Maia 200 rack-scale token throughput; multi-gigawatt deployments planned from 2027 (Microsoft blog, Oct 12 2026; Azure blog; AI Weekly; 288 GB is the capacity reported for eight HBM4 stacks, not stated in Microsoft’s post). Microsoft says it will become the primary inference system for its products, with production deployments starting in 2027.
Timeline
Every dated event in the wiki that involves Microsoft Maia 300, newest first.
New rack designs and accelerators aimed at agentic inference.
Connections
Microsoft Maia 300 and everything one link away. Drag to move, scroll to zoom, click a node to open it.Open in the full graph →
What links here
Concept Inference
Company Microsoft
Sources
- Microsoft, “Microsoft to expand Azure Maia system to multi-gigawatt deployments”, Oct 12 2026
- Azure blog, “Scaling Azure infrastructure through open collaboration”, Oct 13 2026
- AI Weekly, Oct 12 2026 [link needed]