Glossary
Matrix Math Acceleration (MMA)
An on-chip acceleration feature in IBM Power10 and Power11 processors that speeds up the matrix math operations used in AI inferencing, without requiring a separate GPU.
Matrix Math Acceleration is an on-chip Power10 and Power11 processor feature, distinct from a PCIe-attached Spyre Accelerator card. IBM's July 2026 Power S1112 announcement describes local inference using Power 11 MMA. An IBM i shop can evaluate a supported inference runtime on the same physical Power system as production IBM i partitions, but must verify operating-system support, model framework, available cores and memory, and production headroom. The on-chip feature does not mean every model runs natively inside IBM i.