Microsoft Unveils Maia 300 AI Chip to Scale Multi-Gigawatt Internal Deployments

Microsoft has unveiled its third-generation Maia 300 artificial intelligence processor, planning multi-gigawatt data center deployments to serve as the primary inference engine for internal workloads such as Copilot and GitHub. The company aims to curb computing costs and dependence on merchant suppliers like Nvidia amid projected annual capital spending above $175 billion. Microsoft claims Maia 300 delivers 40% more tokens per megawatt than leading GPU architectures, with deployments scheduled between 2027 and next calendar year alongside design partner Marvell Technology, while continuing to buy third-party GPUs for training and external cloud customers.