← Back to live feed1 story
Microsoft plans to deploy over 1 million Maia 300 AI accelerators next year, making its in-house chips the primary system for inference across all services and products including Copilot. The Maia 300, manufactured by TSMC on a 2nm process, features 288 GB of HBM4 memory and claims 40% more tokens per MW than leading GPU-based systems. The company aims to scale its Maia family to multi-gigawatt deployments starting in 2027 to reduce dependence on Nvidia and AMD.
Sign in to suggest
Key sources
- SOURCE@satyanadella“scale our Maia family to multi-gigawatt deployments, making it the primary inference system for our own product and services”x.com
- SOURCE@microsoft“up to 7x the rack-scale token throughput of Maia 200”x.com
- SUPPORT@aaronpholmes“announced Maia 300 today and said it's planning "multi-gigawatt deployments" next year”x.com
- SUPPORT@southernvalue95“expect it to get to the majority of our internal consumption”x.com
- SUPPORT@business“with the aim of powering its AI software using mostly homegrown hardware”x.com
- SUPPORT@singularityres“Reuters says Microsoft plans to deploy potentially more than a million of them next year”x.com
- SUPPORT@thetranscript_“Starting next calendar year, we will scale our Maia family to multi-gigawatt deployments”x.com
- SUPPORT@edzitron“microsoft says a lot of things lol”x.com