Microsoft is preparing a major expansion of its internally developed artificial intelligence hardware, announcing the next generation of its custom Maia chip and plans to potentially deploy more than 1 million units next year. The move reflects a broader shift among major technology companies toward proprietary AI processors as rapidly increasing computing demand makes infrastructure efficiency and operating costs increasingly important.

Microsoft Targets Large-Scale Maia 300 Deployment

The new Maia 300 is the third generation of Microsoft’s custom AI chips, designed to handle intensive computing workloads such as coding and complex business tasks carried out on behalf of users. Microsoft said it could deploy more than 1 million of the chips next year to support its internal AI services, signaling that proprietary silicon is moving from a specialized development effort toward a significant component of the company’s AI infrastructure strategy.

The scale of the planned deployment is important because AI services require substantial computing capacity. Rather than relying exclusively on commercially available processors, Microsoft is developing hardware tailored to the specific requirements of its AI systems. The strategy could give the company greater control over infrastructure design and performance as demand for AI computing continues to expand.

Energy Efficiency Becomes a Core AI Metric

Microsoft said Maia 300 delivers approximately 40% more computing per megawatt than leading GPU-based systems. That efficiency claim is particularly relevant as data centers face rising electricity requirements from increasingly sophisticated AI workloads.

According to Microsoft, Maia 300 racks contain 96 chips and consume approximately 167 kilowatts of power. These specifications illustrate the scale of infrastructure required to operate advanced AI systems. At large deployments, even incremental improvements in computing efficiency can have significant implications for the amount of power and data-center capacity required to deliver AI services.

The emphasis on performance per megawatt also shows how competition in AI infrastructure is evolving. Computing speed remains important, but the cost and availability of electricity, cooling and data-center capacity are becoming increasingly central to the economics of large-scale AI deployment.

Custom Silicon Expands Microsoft’s AI Infrastructure Strategy

Microsoft’s previous Maia chip generation is already running the company’s in-house AI models, as well as models from OpenAI delivered through its Azure cloud computing service. The Maia program therefore forms part of a broader infrastructure strategy connecting Microsoft’s internal AI development with its cloud business.

For Microsoft, expanding custom-chip deployment could provide greater flexibility as AI workloads diversify across software development, business automation and other applications. It also creates a potential avenue to manage the infrastructure costs associated with the rapid expansion of AI services while reducing reliance on a single category of external hardware.

Investors will be watching the rollout of Maia 300, including whether Microsoft reaches its planned deployment scale and whether the reported efficiency advantage translates into meaningful operating benefits. The company’s progress will also offer a broader indication of how quickly custom AI silicon can become an important part of the economics of large-scale artificial intelligence infrastructure.