Microsoft has introduced Maia 200, its first custom accelerator built specifically for large-scale AI inference.
Designed to improve tokens per dollar and per watt, Maia 200 uses narrow-precision compute, software-managed SRAM, and Ethernet-based scale-up to support next-generation AI services.
#AIInfrastructure #DataCenter #EnterpriseIT
https://www.instagram.com/p/DT_VsevEvPF
Posted on Instagram ·
942
5
942
5
Suggested Credits
Credits on Gondola describe each credited role and may be incomplete. A brand or featured person is not necessarily the creator.