New Delhi: Microsoft has unveiled Maia 300, its third-generation in-house artificial intelligence chip, as the company looks to reduce its dependence on processors from Nvidia and AMD and improve the efficiency of running AI services.
The new chip is expected to begin production deployment in 2027. Microsoft plans to expand its Maia chip infrastructure to multi-gigawatt capacity from next year, with the aim of making its own hardware the primary system for handling AI requests across its products and services.
Microsoft CEO Satya Nadella announced the development on X, describing it as the company’s biggest investment yet in its own silicon technology. He said Maia 300 could deliver 40 per cent more AI tokens per watt than leading GPU-based systems.
The announcement comes as demand for AI computing continues to increase, prompting technology companies to develop specialised processors to improve performance, manage electricity consumption and control the cost of delivering AI services.
Maia 300 aims to improve AI computing efficiency
Microsoft says Maia 300 is designed to deliver higher performance while consuming less power. The company expects the chip to process more AI tokens for every unit of electricity consumed, potentially improving the efficiency of its AI infrastructure.
AI tokens are units of text or data processed by an AI model. A system that generates more tokens using the same amount of power can help reduce the resources required to operate AI applications at scale.
According to Microsoft, Maia 300 is expected to deliver 40 per cent more tokens per watt than leading GPU-based systems. This improvement could allow the company to handle a greater volume of AI requests without a proportional increase in electricity consumption.
The company also says a data centre rack containing 96 Maia 300 accelerators can deliver up to seven times the token throughput of a comparable Maia 200 configuration.
The chip is designed for demanding AI workloads, including complex reasoning, longer conversations and tasks involving AI agents that need to complete multiple steps.
These capabilities are relevant to services that require AI models to process user requests continuously. Improving computing efficiency could help Microsoft support growing demand across its software and cloud platforms.
Chiplet-based design sets Maia 300 apart
Maia 300 uses a chiplet-based design that separates computing, memory control and input-output functions. This approach allows different sections of the processor to use technologies suited to their particular requirements.
The design is part of Microsoft’s effort to improve the performance and efficiency of its own AI hardware. By developing processors specifically for its workloads, the company can optimise its infrastructure for the demands of its products and services.
The Maia family is central to this strategy. Microsoft introduced its first Maia chip in 2023, followed by Maia 200 in January 2026.
According to the company, Maia 200 is already handling production workloads across Microsoft Copilot, GitHub Copilot and Microsoft Foundry. Microsoft has also reported efficiency gains from deploying Maia 200 across the United States, Europe and Australia.
Maia 300 builds on this development as Microsoft prepares to move a larger share of its AI computing onto in-house hardware.
Microsoft will continue using Nvidia and AMD chips
The launch of Maia 300 does not mean Microsoft will stop using processors from Nvidia and AMD.
Instead, the company intends to shift a larger share of AI inference workloads to its own chips. Inference is the process through which a trained AI model generates responses to user requests.
Nvidia and AMD will continue to play a role in Microsoft’s computing infrastructure, particularly in AI model training. Nvidia’s graphics processing units, or GPUs, remain widely used for AI workloads, and Microsoft has relied on them to power cloud services and support AI development.
Developing its own processors gives Microsoft another option as the need for computing capacity grows. It also provides greater flexibility in how the company manages its infrastructure.
Electricity consumption is a major operating expense for large data centres. If Maia 300 delivers the efficiency improvements claimed by Microsoft, the company could handle more AI requests with the same amount of power.
The move is therefore aimed at diversifying Microsoft’s hardware strategy rather than eliminating its relationships with established chipmakers.
Could Maia 300 make Copilot cheaper or faster?
The improved efficiency of Maia 300 could eventually benefit users of Microsoft’s AI services, although the company has not announced direct price cuts.
More efficient chips could reduce the cost of running AI applications and allow Microsoft to process additional requests using the same amount of electricity.
Microsoft has also indicated that parts of Microsoft 365 Copilot could become faster with Maia 300. However, the improvements experienced by users will depend on how the chip is deployed and how individual services are optimised.
The company has not announced a direct reduction in subscription prices for Copilot or other AI products as part of the Maia 300 launch.
For customers, the potential benefits will depend on the extent to which hardware-level improvements translate into faster responses, more capacity and lower service costs.
Maia 300 deployment expected in 2027
Microsoft expects to begin deploying Maia 300 in production in 2027. The company also plans to scale its Maia infrastructure to multi-gigawatt capacity from next calendar year.
The expansion is intended to make the Maia family the primary inference system for Microsoft’s products and services.
Nadella said the company would continue building on efficiency improvements while incorporating innovations from across the wider technology ecosystem.
The planned deployment represents another stage in Microsoft’s effort to develop its own AI computing infrastructure. While Nvidia and AMD will remain important suppliers, Maia 300 could help Microsoft take greater control over how its AI services are powered.
The chip’s real-world impact will become clearer as production deployment begins and Microsoft integrates it into its services. For now, the company is positioning Maia 300 as a major step towards improving AI computing efficiency and managing the growing costs of artificial intelligence.