Microsoft unveils its third-gen in-house AI chip, the Maia 300, built on TSMC's N2P process node and aimed at serving as its "primary AI inference system"

Microsoft's internal AI chips will power its services with 40% more efficiency, cutting costs and boosting performance for AI tasks starting in 2027.

Microsoft has unveiled its third-generation in-house AI chip, the Maia 300, designed as its primary AI inference system. Built using TSMC's N2P process node and a new chiplet architecture, Maia 300 aims to significantly improve the cost and efficiency of running AI workloads. Microsoft claims it will offer 40 percent more tokens per megawatt compared to leading GPU-based systems and up to seven times the rack-scale token throughput of its predecessor, Maia 200. The new chip is engineered to handle advanced reasoning, long-context models, and agentic tasks. The Maia 300 is slated for production rollout starting in 2027, with Microsoft planning multi-gigawatt deployments to power its Azure cloud services and internal AI products, potentially reducing reliance on third-party chips from Nvidia and AMD.

The introduction of Maia 300 signifies a strategic move by Microsoft to gain greater control over its AI infrastructure, aiming to optimize costs and performance beyond what off-the-shelf components can offer. This vertical integration allows Microsoft to co-design hardware and software, tailoring chips to the specific demands of its evolving AI models and services like Copilot and its own large language models. The move also highlights a broader industry trend of major tech players developing custom silicon to achieve competitive advantages in the AI race.

By investing heavily in in-house AI chip development, Microsoft is positioning itself to potentially reduce its dependency on Nvidia, a move that could reshape the supply chain dynamics for AI hardware. The projected efficiency gains and increased throughput of Maia 300 suggest a future where cloud providers can offer AI services at more competitive price points or with superior performance. Enterprises relying on Azure for their AI workloads may benefit from these advancements through improved service reliability and cost savings.