Microsoft to Scale Azure Maia to Multi-Gigawatt Deployments From 2027

Microsoft says it will scale its Azure Maia family to multi-gigawatt deployments starting in 2027, with the second-generation Maia 200 already serving production workloads and the third-generation Maia 300 planned for production deployments the same year. The company detailed the roadmap in a blog post by Scott Guthrie, Executive Vice President, Cloud + AI.

Maia 200 is now running Microsoft production workloads including Microsoft Copilot, and is supporting the MAI team as it develops next-generation models. GitHub Copilot and Microsoft Foundry are to follow by the end of the year, according to Microsoft. The chip runs OpenAI models, Microsoft’s own MAI models and leading open-source model families.

What Microsoft says Maia 200 delivers today

The first production generation is deploying across the United States, Europe and Australia. Microsoft claims it delivers up to 1.7x more tokens per watt for built-in AI capabilities in Microsoft Teams, Copilot and Cowork experiences compared with leading GPU infrastructure.

If you use Copilot or Teams, that efficiency claim is the part that touches you directly: the same power budget is meant to serve more AI interactions. Microsoft frames the Maia program as co-designed hardware and software, with insights from running its own AI services feeding back into chip design.

Maia 300 targets 40% more tokens per megawatt

Maia 300 is described as a complete intelligent system built for long-context reasoning, latency-sensitive conversational experiences and sustained multi-step agentic tasks. Microsoft expects it to deliver 40% more tokens per megawatt compared with leading GPU-based systems, and up to 7x the rack-scale token throughput of Maia 200 in a full 96-accelerator configuration.

The company lists several architectural details for the system:

• Eight HBM4 stacks providing 18.0 TB/s of peak memory bandwidth within a 950-watt TDP.
• A multi-chiplet architecture separating compute, memory-controller and I/O functions, with compute chiplets built on TSMC’s N2P process node.
• Support for advanced low-precision formats including enhanced MXFP4, aimed at larger context windows and better model efficiency in the same power envelope.
• Rack systems combining 96 accelerators in a 167-kilowatt power envelope, with 27.6 TB of HBM capacity and 1.7 PB/s of aggregate HBM bandwidth.
• Ethernet-based scale-up networking designed to extend from a single rack to datacenter-scale deployments.

Microsoft says Maia 300 will begin production deployments in 2027 and that Maia will become the primary inference system for Microsoft products and services. The company also says it plans to share more on its strategy at Microsoft Ignite 2026 in San Francisco next month.

For now, there is nothing to install or buy: Maia is Microsoft’s own datacenter silicon, not a product sold to customers. The practical way to follow the rollout is through the Microsoft services that run on it — Copilot, Teams, GitHub Copilot and Microsoft Foundry — and through the technical paper Microsoft published alongside the announcement.

Related reading

Sources

• Microsoft to expand Azure Maia system to multi-gigawatt deployments — Microsoft Blog