Microsoft is set to deploy its proprietary Azure Maia AI chips at a multi-gigawatt scale across its product lineup starting in 2027, according to The Decoder. This marks a significant shift as the company moves to make these in-house chips the primary AI inference system powering its services.
The upcoming Maia 300 chip is expected to be 40 percent more efficient in processing tokens per megawatt compared to leading GPU systems, offering improved performance and energy efficiency. CEO Satya Nadella described this initiative as "our biggest bet yet on our own silicon," highlighting Microsoft's commitment to custom AI hardware development.
For Japanese investors and tech markets, Microsoft's move underscores a growing trend of major tech firms investing in bespoke AI chips to optimize performance and reduce operational costs, a development likely to influence hardware and cloud service strategies regionally.