Washington: Microsoft is preparing to introduce its next generation artificial intelligence chip, Maia 300, as the company works to strengthen its own hardware capabilities and reduce its dependence on outside chip suppliers. The new chip could be unveiled as early as September, marking another important step in Microsoft's effort to build more of the technology needed to power its growing artificial intelligence services.
The planned Maia 300 launch comes at a time when demand for computing power is rising rapidly. Artificial intelligence systems require huge amounts of processing capacity, particularly as companies use more advanced models and increasingly rely on AI assistants and other services. For Microsoft, which operates one of the world's largest cloud computing businesses, controlling the supply and cost of these chips has become an increasingly important part of its strategy.
Microsoft began developing its own Maia family of artificial intelligence chips several years ago. The company introduced its first Maia chip in 2023 before launching Maia 200 in January 2026. Maia 200 was designed mainly for AI inference, which means the process of using trained AI models to produce answers, generate content and perform other tasks. Microsoft said the chip was built to improve speed, energy efficiency and the cost of running AI workloads through its Azure cloud platform.
Maia 300 is expected to take this programme to a much larger scale. Microsoft has been discussing manufacturing capacity with Taiwan Semiconductor Manufacturing Company for more than 300,000 Maia 300 chips, with deliveries reportedly planned for 2027. The company is also reported to have a longer term ambition to secure capacity for more than one million chips.
However, the reported production figures should be treated carefully. Microsoft has pushed back against the specific numbers and said they do not accurately represent the full scale of its custom silicon programme. The company has not publicly confirmed the final production volume, detailed specifications or exact launch date for Maia 300.
The push into custom chips is important because Microsoft's artificial intelligence infrastructure currently depends heavily on chips from major suppliers. Building its own accelerators gives the company another option as the cost and availability of advanced AI hardware become increasingly important.
The strategy does not mean Microsoft is abandoning other chip suppliers. The company continues to use hardware from different manufacturers and is building a mixed infrastructure designed to support different types of AI workloads. Its own chips can instead give Microsoft greater control over specific workloads where customised hardware may offer advantages in performance, energy use or operating costs.
Microsoft is also facing strong competition from other major technology companies that have been developing their own AI processors. Google has its Tensor Processing Units, while Amazon has developed its own Trainium and Inferentia processors. Both companies have moved aggressively to reduce their reliance on third party hardware for some cloud based AI workloads.
Microsoft entered this competition later than some rivals, but the company is now increasing its efforts. The scale being considered for Maia 300 suggests that Microsoft wants its custom silicon to become a significant part of Azure's future infrastructure rather than remaining a small experimental project.
Another potentially important development is Microsoft's interest in attracting outside customers to its Maia based computing systems. Companies developing large AI models need enormous amounts of computing power, and cloud providers are competing to offer them cheaper and more efficient alternatives.
Reports have indicated that Anthropic has explored using Microsoft's Maia chips through Azure. If major AI companies eventually run significant workloads on Maia, it could provide Microsoft with valuable evidence that its chips can compete beyond its own internal systems.
The challenge will be making Maia work efficiently at very large scale. Designing an AI chip is only one part of the problem. Microsoft also needs the software, networking, data centre systems and developer tools required to make the hardware useful for demanding AI applications.
This is one area where established chip companies have a major advantage. Their hardware has been supported by mature software ecosystems and years of development. Microsoft therefore has to ensure that its own chips are not only powerful but also practical and easy to use.
The Maia 300 programme also reflects a wider change in the artificial intelligence industry. Technology companies are increasingly trying to control more parts of the AI infrastructure, from chips and servers to cloud platforms and AI models. The reason is simple. As AI usage grows, computing costs can become one of the largest expenses involved in delivering these services.
For Microsoft, developing Maia could eventually provide greater control over those costs while allowing its engineers to design hardware specifically for the workloads running across Azure.
The expected Maia 300 unveiling will therefore be closely watched. Its technical specifications, performance and production plans will reveal how far Microsoft has progressed in its effort to build a serious alternative source of AI computing power.
The company does not need Maia 300 to completely replace other AI chips for the strategy to succeed. Even capturing a significant share of Microsoft's own enormous computing requirements could make the project strategically valuable.
For now, the most important development is the scale of Microsoft's ambition. Maia 300 appears to be moving the company beyond simply experimenting with custom AI chips and toward building a much larger in house computing platform. As artificial intelligence becomes increasingly central to cloud computing, Microsoft's ability to control the hardware behind that technology could become an important part of its competition in the years ahead.