Key Takeaways:

Microsoft stated that its Maia series AI chips will be deployed at a scale of several gigawatts of computing power in the future.

The company plans to make Maia its primary chip architecture for running AI models.

The next-generation Maia 300 chip will enter production in 2027.

Azure Maia 300 chip prototype

After years of relying on NVIDIA (NVDA) graphics processors, Microsoft (MSFT) is expanding its use of in-house artificial intelligence chips, as the software giant aims to strengthen its full-stack AI capabilities.

Microsoft officially unveiled its third-generation Maia AI accelerator chip on Monday, stating that it will primarily rely on its own chip hardware for handling business requests going forward.

Microsoft's biggest advantage in AI is the Azure cloud platform, which runs various AI models for customers. Microsoft 365 Copilot and GitHub Copilot provide AI services for programmers and office workers. Microsoft is also developing its own large models, though they have yet to gain significant influence in a market dominated by OpenAI, Anthropic, and Google.

The chip race is welcoming new opportunities, driven by sustained strong demand for high-end processors and the market's desire for diversified chip options. NVIDIA (NVDA) GPUs have long been the industry standard for AI computing workloads and model training, and Microsoft's existing data centers extensively use these products, also to run OpenAI and Anthropic models.

However, as demand continues to climb, leading model companies are beginning to adopt chips from multiple vendors, including Amazon, AMD, Cerebras, and Google, and in recent months, Microsoft's in-house chips as well. In general-purpose computing scenarios, Microsoft already uses its own CPUs alongside chips from Intel and AMD.

Chief Financial Officer Amy Hood said at a Goldman Sachs conference last month that in-house chips have helped improve the company's profit margins over the past decade. Microsoft's first AI chip, Maia 100, was launched in 2023, followed by Maia 200. CEO Satya Nadella told analysts on an earnings call that Maia 200 delivers "30% more performance per dollar invested than existing hardware," and the chip is already being used to run OpenAI models.

Maia 300 will begin mass production in 2027. Microsoft Executive Vice President Scott Guthrie said in a blog post that the goal is to achieve gigawatt-scale deployment of the Maia series chips, though General Manager Alistair Speirs did not provide a specific timeline. Speirs said: "We retain some flexibility. This industry iterates extremely quickly, and we want to match the right chip to the right model at the right time."

Microsoft plans to deploy 96 Maia 300 chips per data center rack, double that of Maia 200. Single-rack AI token processing throughput can reach up to 7 times that of the previous generation; compared to GPUs, tokens produced per megawatt of energy consumption increase by 40%. "We need a new architecture to achieve better performance while improving power efficiency. For data centers, power consumption directly corresponds to operating costs," Speirs said.

Microsoft Technical Fellow Jim Kleewein said some Microsoft 365 Copilot operations can achieve speed improvements with Maia 300. Investment bank Evercore ISI analysts estimated in a research note sent to clients on Wednesday that by the first quarter of 2027, Maia chips will account for 10% of Microsoft's AI chip procurement spending; the firm maintains a buy rating on Microsoft stock.

NVIDIA (NVDA) will remain an important partner for Microsoft. Last week, Microsoft held a product launch event, with NVIDIA CEO Jensen Huang in attendance, where Microsoft unveiled new Surface laptops equipped with NVIDIA processors. Speirs also mentioned that Microsoft will continue to collaborate with NVIDIA and AMD on model training business.

Unlike Google's TPUs and Amazon's Inferentia and Trainium chips, Microsoft's accelerator chip is not yet available for hourly rental to other companies. Patrick Moorhead, CEO of industry consulting firm Moor Insights & Strategy, said: "Currently, no external company outside of Microsoft has tested this chip."