# Microsoft Unveils Maia 300 AI Chip, Plans to Run Its Own AI Products Primarily on In-House Silicon

*2026-10-15 22:00 UTC · Technology · 1 min read*

> Microsoft says its third-generation Maia 300 inference accelerator delivers 40% more tokens per megawatt than leading GPU systems, with large-scale deployment starting in 2027.

Microsoft unveiled Maia 300 on October 12, its third-generation in-house AI accelerator, built mainly for inference. The company plans to scale the Maia family to multi-gigawatt deployments from 2027 and make it the primary inference system behind its own products and services.

Microsoft claims Maia 300 delivers 40% more tokens per megawatt than leading GPU-based systems, and up to seven times the rack-scale throughput of Maia 200 in a full 96-accelerator rack. The chip reportedly uses a chiplet design on TSMC's N2P process with HBM4 memory. Microsoft has not named the GPU systems it compared against, and no independent benchmarks have been published.

Maia 200 already runs Copilot and other Microsoft cloud services, but the company still relies heavily on Nvidia and AMD GPUs. Reuters reportedly said deployment could mean more than one million Maia 300 chips next year.

## Sources

- [itdaily.com](https://itdaily.com/news/datacenter/microsoft-maia-300/)
- [datacenterknowledge.com](https://www.datacenterknowledge.com/data-center-chips/microsoft-pushes-custom-ai-chips-with-azure-maia-300)

---
TechTools News · https://techtools.cz/tools/news/article/2524-microsoft-unveils-maia-300-ai-chip-plans-to-run-its-own-ai-products-primarily
Česká verze: https://techtools.cz/tools/news/clanek/2598-microsoft-predstavil-ai-cip-maia-300-a-chce-sve-ai-produkty-provozovat-hlavne