Microsoft: Maia 300 delivers 40% more tokens per watt; Maia to be primary inference silicon from next year
What's the bet? Microsoft CEO Satya NadellaDealroom has a profile for this one. Try Dealroom → said the company is placing its biggest bet yet on its own silicon. From next calendar year it will scale the Maia family to multi-gigawatt deployments as the primary inference system for Microsoft’s own products and services, and is introducing Maia 300.
Performance claim: Nadella said Maia 300 delivers 40% more tokens per watt compared with leading GPU-based systems. Maia 200 is already deploying across the US, Europe and Australia, delivering up to 1.7x more tokens per watt across Teams and Copilot experiences. The goal is to build on full-stack efficiency gains while still bringing in the latest innovations from across the ecosystem.
Who amplified it: Microsoft AI CEO Mustafa SuleymanDealroom has a profile for this one. Try Dealroom → quote-posted the announcement, highlighting the 40% tokens-per-watt gain.
Source: Satya Nadella on X
