Microsoft Plans to Reveal Custom Artificial Intelligence Chip to Cut Cloud Costs
Redmond, Monday, 10 August 2026.
Microsoft plans to unveil its Maia 300 artificial intelligence chip in September 2026, while negotiating with TSMC to produce over 300,000 units to reduce reliance on Nvidia.
Microsoft Schedules Maia 300 Unveil for September 2026
Microsoft Corporation (NASDAQ: MSFT) is preparing to officially introduce its next-generation Maia 300 artificial intelligence silicon this fall, with reports indicating a potential unveiling as early as September 2026 [1][2]. This strategic move highlights the tech giant’s ongoing effort to expand its in-house chip infrastructure, aiming to optimize cloud performance for complex AI workloads and reduce long-term operational reliance on third-party semiconductor vendors like Nvidia [1][3]. The company originally introduced its Maia AI chip in November 2023 but has since sought to accelerate development to catch up with rivals such as Alphabet in scaling proprietary silicon [2][3]. Today, on 10 August 2026, the anticipation builds for the formal announcement expected within the next month [1].
Production Targets and Supply Chain Negotiations
To support this initiative, Microsoft is currently negotiating with Taiwan Semiconductor Manufacturing Company (TSMC) to secure production capacity for over 300,000 units, with delivery scheduled for 2027 [4][6]. While the current Maia 200 chips have been produced in the tens of thousands to date, the company ultimately wants capacity for over 1,000,000 chips pending supply chain resolution [5][6]. Andrew Wall, General Manager of Azure Maia, has indicated that the Maia 300 is expected sometime in 2027, though the Maia 200 is expected to remain in use for another four to five years [4]. These production figures remain under negotiation and are not yet signed purchase orders, reflecting the complex nature of semiconductor supply chains [6].
Market Strategy and Cost Efficiency Goals
The strategic objective for the Maia 300 is to reduce inference costs for internal models and OpenAI workloads, serving as an alternative to reliance on Nvidia GPUs currently anchoring Azure infrastructure [6]. Reports suggest that current Maia 200 chips are already 30%-40% cheaper to operate than cutting-edge Nvidia chips for specific workloads, a metric the new silicon aims to improve upon [5]. Microsoft is targeting major cloud customers like Anthropic to use the chip, despite Anthropic concurrently developing its own internal semiconductor team [3][6]. This effort represents a significant shift towards homegrown AI accelerators, moving beyond a hedge against Nvidia to a serious replacement strategy [6].