Microsoft unveils Azure Maia 300 AI accelerator system

Microsoft has announced Azure Maia 300, a third-generation AI accelerator system optimized for LLM inference and cloud-scale agentic workloads.

• The system is co-designed across software, silicon, rack infrastructure, networking, and datacenters, and is intended to maximize tokens per watt and tokens per dollar for Microsoft’s AI products.
• Microsoft says Maia 300 delivers up to 7x greater rack-scale token throughput and more than 2.5x performance per watt versus Maia 200; the design includes 96 accelerators per rack, 12,288 accelerators per superpod, 167 kW racks, and 8 HBM4 stacks with 288 GB capacity and 18 TB/s bandwidth per accelerator.