Custom Silicon at the Core of the Expansion

The multi-gigawatt buildout will be anchored by Microsoft's Maia accelerator family. The company's blog materials include specification imagery for the Maia 300, the latest chip in the line, alongside product visuals for the Maia 200. Microsoft positions the custom silicon as the foundation for running large-scale AI workloads across its Azure cloud infrastructure.

The scale of the expansion — measured in gigawatts of power capacity rather than individual chips or racks — underscores the energy demands of next-generation AI datacenters. Microsoft did not detail the specific number of accelerators or total capital expenditure tied to the program in the announcement.

Timeline Set for 2027

The expansion to multi-gigawatt operation is scheduled to start in 2027, meaning the larger-scale deployment remains ahead rather than a description of current capacity. Until then, Microsoft's existing Maia-based infrastructure continues to operate at a smaller footprint while the company prepares the 2027 buildout.

No Korean-language reporting was used in the preparation of this item; the announcement is based on Microsoft's primary English-language materials.

The claim that Microsoft is expanding its Azure Maia AI system to multi-gigawatt scale, with the expansion beginning in 2027, is True.