API
< Back to all clusters
[TECHNOLOGY] · United States, Taiwan · 87 sources

started · updated

Microsoft unveils Maia 300 AI chip to boost inference efficiency

Microsoft has unveiled the Maia 300, the third generation of its custom-designed AI accelerator chip, intended to serve as the primary inference system for its Azure cloud services and various products. The new chip is designed to handle advanced AI workloads, including long-context reasoning, latency-sensitive conversational experiences, and multi-step agentic tasks.

Built using a modular chiplet architecture, the Maia 300 utilizes TSMC’s N2P process for its central compute core, while employing a 3nm process for input-output and memory controller functions to optimize costs. The system features HBM4 memory and is designed for high efficiency, with Microsoft claiming it will deliver 40% more tokens per megawatt than leading GPU-based systems. In a full 96-accelerator rack configuration, the Maia 300 is expected to provide up to seven times the token throughput of the previous Maia 200 generation.

Microsoft plans to begin production deployment of the Maia 300 in 2027, with ambitions to scale the Azure Maia family to multi-gigawatt deployments. This strategic move aims to reduce the company’s reliance on third-party hardware providers such as Nvidia and AMD while improving the cost-effectiveness and energy efficiency of its AI infrastructure.

Entities

AMD · Azure · Google · Maia 300 · Microsoft · Nvidia · Satya Nadella · Scott Guthrie · TSMC

Claims

What the coverage asserts, and how many sources carry each claim.

Sources

科技速递:微软CEO称AI是潜在内部威胁,月之暗面增资至约1437万_天极网 [yesky.com]