Microsoft said in an Oct. 12 blog post that it will scale its Azure Maia family to multi-gigawatt deployments starting in 2027. Maia is Microsoft's own line of AI accelerator systems. The plan would make it "the primary inference system for Microsoft products and services." Maia 300, the next generation, is scheduled to begin production deployments in 2027. The chip in service now is its predecessor. Maia 200 is already serving Microsoft production workloads, including Microsoft Copilot.
Key facts from the Microsoft announcement
- In production now: Maia 200 serves Microsoft production workloads, including Microsoft Copilot.
- Next on Maia 200: GitHub Copilot and Microsoft Foundry are to follow by end of year.
- Models supported: OpenAI models, Microsoft's own MAI models and leading open-source model families run on Maia 200.
- Next generation: Maia 300 production deployments are planned to begin in 2027.
- Scale: Multi-gigawatt Maia deployments are planned from 2027, with Maia as the primary inference system for Microsoft products and services.
What runs on Maia 200 today
Microsoft says Maia 200 is serving production workloads that include Microsoft Copilot. Those workloads run OpenAI models, Microsoft's own MAI models and leading open-source model families. More products are lined up. GitHub Copilot and Microsoft Foundry are to follow on Maia 200 "by end of year." The 2027 dates belong to Maia 300 production deployments and the multi-gigawatt plan. The dated steps run from the Oct. 12 announcement to those plans.
Maia timeline: what runs now and what is planned
Step through what Microsoft says is running on Maia now and what it plans for 2027.
- Oct. 12, 2026Maia 200 serving production workloadsdone
The announcement says Maia 200 is now serving Microsoft production workloads, including Microsoft Copilot. It runs OpenAI models, Microsoft's own MAI models and leading open-source model families.
Source: Microsoft, Oct 12, 2026 - 2027Maia 300 production deployments beginannounced
Microsoft says Maia 300 will begin production deployments in 2027. This is a plan, not a completed rollout.
Source: Microsoft, Oct 12, 2026 - Starting in 2027Multi-gigawatt Maia deploymentsannounced
Microsoft plans to scale the Azure Maia family to multi-gigawatt deployments, making Maia the primary inference system for its products and services.
Source: Microsoft, Oct 12, 2026
The first date is the announcement's publication date, and the later steps are Microsoft's stated plans with no completion confirmed.
What Maia 300 is, in hardware
Microsoft lists the Maia 300 design in the same post. It uses eight HBM4 stacks. They provide 18.0 TB/s of peak memory bandwidth within a 950-watt TDP. A Maia 300 rack brings together 96 accelerators within a 167-kilowatt power envelope. In that full 96-accelerator configuration, Maia 300 delivers up to 7x the rack-scale token throughput of Maia 200.
Microsoft's efficiency comparisons
Microsoft expects Maia 300 to deliver 40% more tokens per megawatt than leading GPU-based systems. The blog post leaves the baseline systems unspecified. A separate Azure Blog deep dive compares Maia 300 with its predecessor instead. It puts Maia 300 at more than 2.5 times the performance per watt of Maia 200. The older chip has its own GPU comparison. Maia 200 delivers up to 1.7x more tokens per watt than leading GPU infrastructure. That figure covers built-in AI capabilities in Microsoft Teams, Copilot and Cowork experiences.
Analysis: why the figures do not stack
Placed side by side, the numbers can mislead. Maia 300 is said to beat Maia 200 by more than 2.5 times per watt. Maia 200 is said to beat leading GPU infrastructure by up to 1.7x. Chaining those two figures does not give Maia 300's lead over GPUs, which Microsoft expects to be 40% more tokens per megawatt. The up to 1.7x figure covers built-in AI capabilities in Teams, Copilot and Cowork experiences. The 40% compares against GPU-based systems that go unnamed, so the baselines may not match. The more than 2.5 times figure compares two Maia generations. The up to 7x figure applies to a full 96-accelerator rack. Each number answers a different question.
Analysis: what the 2027 plan stakes
The 2027 plan is larger than a chip launch. Making Maia the primary inference system would put Microsoft products and services mainly on the company's own hardware. The deployment target is multi-gigawatt. The efficiency claims are given in tokens per watt and per megawatt. The model mix also matters. Maia 200 already runs OpenAI models, Microsoft's MAI models and leading open-source families. That range suggests the hardware is built to carry models from outside Microsoft as well as its own.
Limitations
Microsoft's performance figures come without full benchmark methods in either post, and the "by end of year" target for GitHub Copilot and Microsoft Foundry names no year.
What to watch in 2027
The first test is whether Maia 300 production deployments start in 2027 as scheduled. The second is whether the multi-gigawatt scale-up begins the same year. When results appear, the useful details are the workloads, configurations and benchmark methods behind each number. A separate explainer covers NVIDIA and Cisco's USFP optics plan, another set of vendor claims for AI infrastructure.
Sources: Microsoft, Oct 12, 2026 · Deep dive into Azure Maia 300 architecture | Microsoft Azure Blog (azure.microsoft.com), Oct 12, 2026
ZIPMEX promotes trading on trade.zipmex.com. Promotions are labelled and kept separate from news coverage, which is selected and checked without regard to them. Perpetual futures are leveraged derivatives. Prices can move fast and you can lose all of your margin. Not available in every jurisdiction. Not investment advice.
