Article

Microsoft has launched the Maia 200, an advanced AI inference accelerator featuring over 140 billion transistors and 216GB of HBM3e memory, achieving over 10 petaFLOPS of FP4 performance. Designed to improve AI token generation efficiency by 30% per dollar compared to its predecessor, the Maia 100, it supports multiple AI models, including OpenAI's GPT-5.2. This chip, deployed in Microsoft's Iowa datacenter with plans for expansion to Arizona, aims to reduce reliance on NVIDIA and strengthen Microsoft's AI chip ecosystem.