Microsoft Unveils Custom AI Chips to Power Future Infrastructure

2026-10-12 companies

Redmond, Monday, 12 October 2026.
Microsoft will deploy its custom Maia 300 chips in 2027, cutting computing costs and reducing reliance on external suppliers while boosting AI energy efficiency by 40%.

Strategic Announcement and Timeline

Microsoft Corporation (NASDAQ: MSFT) announced on October 12, 2026, the launch of its third-generation custom AI chip, the Azure Maia 300 [1][5]. The company plans to initiate multi-gigawatt deployments of this infrastructure starting in 2027 [2][5]. This move is designed to make the Maia family the primary inference system for Microsoft products and services [4][7]. The announcement was made via the company’s official blog and news channels on Monday, October 12, 2026 [1][5].

Technical Architecture and Efficiency

The Maia 300 utilizes a chiplet architecture built on TSMC’s 2-nanometer process node for key computing functions [1][8]. In a full configuration of 96 accelerators, the system is expected to deliver up to seven times the token throughput of the predecessor Maia 200 at the server-rack scale [5][8]. Microsoft claims the new chips will achieve 40% more tokens per megawatt compared to leading GPU-based systems [3][5]. Regarding power consumption, the Maia 300 racks use about 167 kilowatts of power, which is 25 kilowatts more than Nvidia’s most recent GB300 AI server racks that use 142 kilowatts [1].

Existing Maia 200 Deployment

The predecessor, Maia 200, is already running Microsoft’s in-house AI models and models from OpenAI delivered on Azure [1][5]. Current deployments of Maia 200 are operational across the United States, Europe, and Australia [5]. This existing infrastructure supports workloads including Microsoft 365 Copilot, with GitHub Copilot and Microsoft Foundry expected to adopt the technology by the end of 2026 [2][5]. The Maia 200 currently delivers up to 1.7 times more tokens per watt compared to alternative GPU-based systems [5][8].

Market Position and Supply Chain

This strategic shift highlights a broader trend among major technology firms seeking vertical integration to lower operational expenditures [1][6]. Hyperscalers including Amazon, Google, and Meta are developing custom chips to reduce dependence on Nvidia and AMD [3][8]. Microsoft aims to secure supply chain autonomy amid rising global demand for compute power [1][3]. Further details on the strategy will be presented at Microsoft Ignite 2026 in San Francisco, scheduled for November 2026 [5].

Sources

Artificial Intelligence Semiconductors