{"data":{"slug":"microsoft-maia-200","etype":"chip","name":"Microsoft Maia 200","cat":"AI Accelerator","subtitle":"Microsoft's 3 nm inference accelerator: 216 GB of HBM3E and over 10 PFLOPS of FP4.","summary":"Maia 200, unveiled January 26, 2026, is Microsoft's second-generation in-house accelerator and is built for inference: more than 140 billion transistors on TSMC 3 nm, over 10 PFLOPS of FP4, 216 GB of HBM3E at 7 TB/s and 272 MB of on-chip SRAM in a 750 W package. It scales to 6,144 accelerators over an Ethernet-based two-tier network. Microsoft says it is in production serving Microsoft 365 Copilot, Foundry, OpenAI GPT models and MAI models, deploying across the US, Europe and Australia. It is used inside Microsoft services rather than sold as a chip, and no public Azure instance had been announced as of November 2026.","attrs":[{"g":"Overview","l":"Designer","v":"Microsoft"},{"b":"Official","g":"Overview","l":"Announced","n":"Microsoft blog","v":"January 26, 2026"},{"b":"Official","g":"Overview","l":"Status","n":"Oct 2026; first region US Central (Iowa), then US West 3 (Arizona); Microsoft blog","v":"In production in Azure for Microsoft services"},{"b":"Vendor spec","g":"Manufacturing","l":"Process","n":"Microsoft blog, Jan 2026","v":"TSMC 3 nm"},{"b":"Vendor spec","g":"Manufacturing","l":"Transistors","n":"Microsoft blog, Jan 2026","v":"Over 140 billion"},{"b":"Vendor spec","g":"Compute","l":"FP4","n":"Per chip; Microsoft blog, Jan 2026","u":"PFLOPS","v":">10"},{"b":"Vendor spec","g":"Compute","l":"FP8","n":"Per chip; Microsoft blog, Jan 2026","u":"PFLOPS","v":">5"},{"b":"Vendor spec","g":"Memory","l":"Memory","n":"Microsoft blog, Jan 2026","v":"216 GB HBM3E"},{"b":"Vendor spec","g":"Memory","l":"Memory bandwidth","n":"Microsoft blog, Jan 2026","u":"TB/s","v":"7"},{"b":"Vendor spec","g":"Memory","l":"On-chip SRAM","n":"Microsoft blog, Jan 2026","v":"272 MB"},{"b":"Vendor spec","g":"Power","l":"SoC TDP","n":"Microsoft blog, Jan 2026","u":"W","v":"750"},{"b":"Vendor spec","g":"Physical","l":"Scale-up","n":"Ethernet-based two-tier network; Microsoft blog, Jan 2026","v":"2.8 TB/s bidirectional per chip; 4 per tray; clusters up to 6,144"},{"b":"Vendor spec","g":"Physical","l":"Cooling","n":"Microsoft blog, Jan 2026","v":"Second-generation closed-loop liquid heat exchanger unit"},{"b":"Vendor claim","g":"Performance","l":"Vendor comparison","n":"Microsoft blog, Jan 2026","v":"~30% better performance per dollar than Microsoft's prior fleet hardware; 3x FP4 of Trainium3"},{"b":"Vendor claim","g":"Performance","l":"Efficiency","n":"Microsoft blog, 2026-10-12","v":"Up to 1.7x tokens per watt vs leading GPU infrastructure (Microsoft services)"},{"b":"Official","g":"Customers","l":"Workloads","n":"Microsoft blogs, Jan and Oct 2026","v":"Microsoft 365 Copilot, Microsoft Foundry, OpenAI GPT-5.2, MAI models; GitHub Copilot by end of 2026"}],"kf":{"node":"TSMC 3 nm","year":2026,"class":"AI Accelerator","tdp_w":750,"mem_bw":7.0,"memory":"216 GB HBM3E","designer":"Microsoft"},"aliases":["Maia 200","Azure Maia 200"],"ticker":null,"exchange":null,"hq":null,"founded":null,"website":null,"status":"Shipping","region":null,"asof":"2026-11-16","sources":[{"url":"https://blogs.microsoft.com/blog/2026/01/26/maia-200-the-ai-accelerator-built-for-inference/","label":"Microsoft: Maia 200, the AI accelerator built for inference (Jan 2026)"},{"url":"https://blogs.microsoft.com/blog/2026/10/12/microsoft-to-expand-azure-maia-system-to-multi-gigawatt-deployments/","label":"Microsoft: Azure Maia to multi-gigawatt deployments (Oct 2026)"}]},"meta":{"found":true,"count":1,"served_from":"site cache (no live upstream calls)","as_of":"2026-11-16T22:03:32+00:00","disclaimer":"Research and information only, not investment advice. Structured records carry basis labels (vendor spec / reported / estimate); market quotes are delayed/indicative.","attribution":"Records compiled from vendor documentation, official filings and reputable public reporting; market quotes via financial data providers. See /data-sources."}}