Microsoft debuts multichiplet Maia 300 inference accelerator

What happened

The company stated the Maia chip series will […] The post Microsoft debuts multichiplet Maia 300 inference accelerator appeared first on SiliconANGLE . UPDATED 16:11 EDT / OCTOBER 12 2026 INFRA Microsoft debuts multichiplet Maia 300 inference accelerator by Maria Deutscher Microsoft Corp. Microsoft is a software and software development company based in Redmond.

today pulled back the curtains on Maia 300, its newest internally developed artificial intelligence accelerator. Microsoft estimates that it can process 40% more tokens per megawatt than “leading GPU (a chip built for many calculations at once, used for graphics and AI)-based systems,” which is likely a reference to Nvidia Corp.

The company stated the Maia chip series will become its primary source of inference capacity over time. Maia 300 features four computing chiplets made using Taiwan Semiconductor Manufacturing Co.’s latest two-nanometer (a label for how small and dense a chip’s circuitry is) node.

The chiplets are flanked by eight HBM4 modules with 288 gigabytes of combined capacity, the same amount as Nvidia Corp.’s flagship Rubin accelerator. However, Maia 300 has about 20% less memory bandwidth than Rubin.

Memory bandwidth is a measure of how much fast data can be transferred to and from RAM. According to Microsoft, Maia 300 features a dataflow architecture designed to minimize data movement.