Stories
Microsoft Set to Scale In-House AI Chips to Multi-Gigawatt Deployments by 2027
- •Microsoft announced plans to scale its Azure Maia AI chip family to multi-gigawatt deployments starting in 2027 to power AI models across its products.
- •CEO Satya Nadella introduced Maia 300, projecting it will deliver 40 percent more tokens per watt than leading GPU-based systems.
- •The expansion follows Maia 200, which is currently deployed in the United States, Europe, and Australia for services like Microsoft Teams and Copilot.
- •Maia 300 is designed for demanding workloads such as long-context reasoning and conversational applications, offering up to seven times the rack-scale token throughput of Maia 200.
- •The new Maia 300 systems will combine 96 accelerators using Taiwan Semiconductor Manufacturing Company’s N2P process technology and are expected to begin production deployments in 2027.
Key Questions Answered
Microsoft announced plans to scale its in-house Azure Maia AI chip family to multi-gigawatt deployments starting in 2027. This initiative aims to make the homegrown silicon the primary system for running artificial intelligence models across the company's products and services.
Microsoft expects the Maia 300 chip to deliver up to 40 per cent more tokens per watt than leading GPU-based systems and up to seven times the rack-scale token throughput of Maia 200 in a full 96-accelerator configuration. The new system is designed to handle demanding workloads like long-context reasoning and conversational applications.
The Maia 200 chip is already deployed across the United States, Europe, and Australia. It is actively serving production workloads including Microsoft Copilot, Microsoft 365 Copilot, and models from OpenAI and open-source communities.
