Microsoft recently introduced Maia 200, their latest custom AI chip, boasting to be the most potent first-party silicon among major cloud providers.

According to Microsoft, Maia 200 outperforms Amazon’s Trainium chip by three times on specific benchmarks and surpasses Google’s latest tensor processing unit (TPU) on various metrics.

The chip is currently operational at Microsoft’s data center near Des Moines, Iowa, where it is supporting OpenAI’s GPT-5.2 models, Microsoft 365 Copilot, and internal projects from the Superintelligence team. A second deployment at a data center near Phoenix is in the works.

Major cloud players are increasingly developing their custom AI chips instead of relying solely on Nvidia. Google has been enhancing its TPUs for almost a decade, while Amazon’s Trainium line is now in its third generation, with a fourth already announced.

Microsoft initially revealed its custom chip ambitions in late 2023 with Maia 100 at the Ignite conference. Despite being a late entrant, Microsoft emphasizes the advantage of tight integration between chips, AI models, and applications like Copilot.

Maia 200 offers 30% better performance-per-dollar compared to existing hardware and focuses more on inference, the process of executing AI models post-training, building on the first-generation chip.

Competition among cloud giants in the chip arena has intensified as the cost of running AI models becomes a significant concern. While training a model is a one-time expense, serving it to millions of users is an ongoing challenge. Custom chips tailored to specific workloads are believed to be more cost-effective than relying solely on Nvidia.

Microsoft is also extending opportunities to external developers by launching a software development kit to optimize models for Maia 200. Interested developers and researchers can sign up for an early preview today.