Microsoft has announced that come next calendar year, it plans on rolling out Azure Maia, its scalable AI infrastructure across multi-gigawatt deployments. This move means that Microsoft will effectively be taking on NVIDIA in the space of AI infrastructure.
“Today we are announcing our biggest bet yet on our own silicon. Starting next calendar year, we will scale our Maia family to multi-gigawatt deployments, making it the primary inference system for our own product and services. And we’re introducing Maia 300, which delivers 40% more tokens per watt compared to leading GPU-based systems, “Satya Nadella, CEO of Microsoft, posted on X. “We are already deploying Maia 200 across the US, Europe, and Australia, delivering up to 1.7x more tokens per watt across experiences in Teams and Copilot. Our goal is to build on the full-stack efficiency gains we’ve already seen, while continuing to bring the latest innovations from across the ecosystem into our fleet.”
To break it down, Microsoft’s Azure Maia scalable infrastructure can be divided into two fronts: Maia 200, and Maia 300.
Maia 200
Maia 200 is designed for real-world workloads that Microsoft runs at scale, which the company claims to be able to deliver up to 1.7x more tokens per watt for built-in AI capabilities, usually for the company’s Teams, Copilot, and Cowork apps
Maia 300
The third generation of Microsoft’s Azure Maia program, this is designed with an inference-first architecture, optimised for long-context, memory-intensive, and decode-heavy AI workloads. This involves balancing compute, memory bandwidth, networking, and power efficiency at cloud scale.
Some key capabilities of the Maia 300 system include less data movement and more useful compute; higher memory bandwidth within every watt; a new multi-chiplet architecture based on TSMC’s N2P process node; and scalable networking at datacentre scale, with a higher AI throughput from every rack.
(Source: Microsoft)
