Microsoft Announces Multi-Gigawatt Scale-Up of In-House Maia AI Chips Starting 2027
Microsoft confirms it will deploy its custom Azure Maia AI accelerators at multi-gigawatt scale from 2027, aiming to reduce reliance on third-party hardware and improve inference efficiency.
ImportanceStructuralEvidenceE3 inspectableWrite-upStandard
Microsoft announced on October 12 that it will scale its in-house Azure Maia family of AI accelerators to multi-gigawatt deployments starting in 2027, making Maia the primary inference system for its products and services.
The second-generation chip, Maia 200, is currently running in production environments across North America, Europe, and Australia, supporting workloads such as Microsoft Copilot. The company states that Maia 200 delivers up to 1.7 times more tokens per watt for Microsoft Teams and Copilot experiences compared to leading GPU infrastructure.
The upcoming third-generation chip, Maia 300, features a TSMC N2P process node and HBM4 memory stacks. Microsoft projects that Maia 300 will deliver 40% more tokens per megawatt than leading GPU-based systems, and up to 7 times the rack-scale token throughput of Maia 200 in a full 96-accelerator configuration.
This move reflects a broader strategic pivot among major hyperscalers toward vertically integrated hardware designs. Google, Amazon, and Meta have similarly accelerated proprietary chip programs to handle specialized AI calculations, challenging Nvidia's historic dominance over the datacenter accelerator market.