Microsoft makes its Maia chips the primary inference system for its own products from next year
Satya Nadella said Microsoft will scale its Maia family to multi-gigawatt deployments starting next calendar year as the primary inference system for its own products and services. He said Maia 300 delivers 40% more tokens per watt than leading GPU-based systems. Chamath Palihapitiya read it as everyone competing with everyone in AI; @jpinsights called it bullish for AVGO and MRVL.