Microsoft unveils Maia 300 for long-context AI agents

What is changing?

Microsoft announced its third-generation Azure Maia 300 AI accelerator on October 12. The company says the system is designed for long-context reasoning, latency-sensitive conversations, and sustained multi-step agentic tasks. Maia 300 uses a chiplet architecture that separates compute, memory-controller, and input-output functions.

Each accelerator provides 288 GB of memory across eight HBM4 stacks and 18 TB/s of peak memory bandwidth. According to Microsoft, a 167 kW rack with 96 accelerators can deliver up to seven times Maia 200’s rack-scale token throughput and more than 2.5 times its performance per watt. The company also expects 40 percent more tokens per megawatt than leading GPU-based systems, but did not identify the GPUs used for that comparison.

Production deployments of Maia 300 are scheduled to begin in 2027. Microsoft plans to scale the Maia family to multi-gigawatt deployments that year and make it the primary inference system across its products and services. Maia 200 is already serving production workloads including Copilot.