Microsoft Plans Maia 300 Chip to Cut Nvidia Reliance

What happened

Microsoft is preparing to unveil its Maia 300 accelerator as early as September, targeting delivery of more than 300,000 units in 2027. The company is also discussing manufacturing capacity for more than 300,000 chips with Taiwan Semiconductor Manufacturing (TSMC). Microsoft introduced Maia in 2023 and deployed Maia 200 in January, an inference-focused chip on TSMC's 3-nanometer process that delivers 30% better performance per dollar than Microsoft's previous generation hardware.

Why it matters

Azure and other cloud-services revenue jumped 40% year over year in Microsoft's fiscal third quarter, with Microsoft Cloud revenue reaching $54.5 billion. Deploying hundreds of thousands of internally designed accelerators could reduce dependence on Nvidia, diversify Microsoft's compute supply, and potentially lower the cost of serving increasingly intensive AI workloads—directly improving the economics of Azure's AI operations.

What to watch

Whether Maia 300 achieves strong performance-per-dollar, whether TSMC successfully ramps production, and whether outside customers such as Anthropic actually adopt the chip. A successful 2027 ramp would strengthen Microsoft's case for capturing more of the economics underlying Azure AI; production delays, weak adoption, or disappointing performance would leave its costly reliance on third-party accelerators largely intact.