TL;DR: Microsoft is making custom silicon its primary inference platform, claiming Maia 300 delivers 40% more tokens per watt than leading GPU-based systems.
Summary: Satya Nadella announced that Microsoft will scale its Maia family to multi-gigawatt deployments starting next calendar year, making it the primary inference system for its own products and services. The new Maia 300 is claimed to deliver 40% more tokens per watt than leading GPU-based systems, while the current Maia 200 is already deployed across the US, Europe, and Australia at up to 1.7x more tokens per watt in Teams and Copilot experiences. Microsoft frames the move around full-stack efficiency gains while continuing to integrate innovations from the broader ecosystem.
Why it matters: Hyperscaler inference shifting to in-house accelerators signals growing pressure on GPU pricing and supply for AI builders. Watch for Maia 300 availability and whether Microsoft exposes comparable efficiency or cost tiers to Azure customers.
Source: x_com