Microsoft is preparing to unveil its next-generation Maia 300 artificial intelligence chip as early as September, according to a recent report. The new processor represents the company's latest effort to expand its custom silicon portfolio for AI workloads hosted in its Azure cloud infrastructure, an area that has taken on growing strategic importance as generative AI adoption drives unprecedented demand for compute across the industry.

The Maia 300 is expected to succeed the Maia 100, which Microsoft first announced at its Ignite conference in November 2023 alongside the Cobalt 100 CPU. The original Maia 100 was designed to accelerate large-language-model training and inference, and the new chip is anticipated to push performance further for these demanding workloads. The Maia 100 was initially positioned to support workloads for Microsoft's deepening infrastructure partnership with OpenAI, and the Maia 300 is expected to broaden that capability across a wider range of Azure-hosted AI services.

According to the report, Microsoft plans to launch the Maia 300 this autumn. The company is reportedly securing production capacity for more than 300,000 chips by 2027, with an eventual target of deploying over one million Maia 300 units. The production is said to be handled by Taiwan Semiconductor Manufacturing Company (TSMC), Microsoft's established foundry partner for its custom silicon program. TSMC's advanced packaging capacity has become a critical bottleneck industry-wide for AI accelerator production, making early foundry commitments a key competitive factor.

The push into proprietary AI accelerators is widely seen as part of Microsoft's strategy to reduce its heavy reliance on Nvidia's processors, which have become the industry standard for training and running large AI models but carry premium price tags. During the peak of the generative AI buildout, Nvidia GPU supply constraints left many cloud providers unable to meet customer demand, underscoring the risk of depending on a single supplier. By developing in-house alternatives, Microsoft aims to lower infrastructure costs and gain more control over its hardware roadmap.

In addition to serving its own internal AI services, Microsoft is also looking to attract major external cloud customers. The report specifically names Anthropic, the AI startup known for its Claude family of large language models, as a potential client for the new chips. Offering custom AI infrastructure could help Microsoft differentiate its Azure cloud platform in an increasingly competitive market for AI computing services.

Microsoft's custom silicon initiative forms part of a broader industry trend. Several major cloud providers, including Amazon Web Services with its Trainium and Inferentia chips and Google Cloud with its Tensor Processing Units (TPUs), have invested heavily in designing their own AI accelerators to reduce dependency on third-party suppliers. Google has been developing and deploying TPUs internally for nearly a decade, giving it a significant head start, while Amazon has steadily expanded its custom chip lineup across multiple generations. Microsoft entered this space more recently but has accelerated its efforts as AI workloads have become central to cloud growth.

Source: Economic Times Markets