Azure Maia 200 and Maia 300
Microsoft says its Maia 200 AI accelerator is in production serving Copilot, and unveiled Maia 300 for 2027 deployment.
Overview
On October 12, 2026, Microsoft said its in-house Maia 200 accelerator is in production, running Copilot and other Azure services in the US, Europe, and Australia, and unveiled the third-generation Maia 300. Microsoft claims Maia 200 delivers up to 1.7x more tokens per watt than leading GPU infrastructure for Teams, Copilot, and Cowork features. GitHub Copilot and Microsoft Foundry are expected to run on it by year-end, and Maia 300 is planned for multi-gigawatt production deployment in 2027. All performance figures are Microsoft's own claims.
This is a sourced directory entry. Features and capabilities are reported in the linked sources; this entry does not document independent Aymorphic testing.
Key features
- Maia 200 in production in the US, Europe, and Australia
- Claimed up to 1.7x tokens per watt versus leading GPU infrastructure
- Runs OpenAI models, Microsoft MAI models, and open-source model families
- GitHub Copilot and Microsoft Foundry expected on Maia 200 by end of 2026
- Maia 300: claimed 40% more tokens per megawatt, up to 7x rack-scale throughput of Maia 200 in a 96-accelerator rack
- Maia 300 production deployment planned for 2027
Use cases
- Serving Microsoft Copilot, Teams, and Cowork inference
- Azure inference capacity alongside Nvidia and AMD GPUs
Pricing
Not sold separately; capacity is consumed through Microsoft services. No pricing published.
Pricing reflects the source record and may change. Confirm current terms with the provider.
Dates and sources
- Released
- Not listed
- Record updated
- 16 Oct 2026
- Record last checked
- 16 Oct 2026