IBM wants its next mainframe generation to speak both instruction languages natively, announcing at Hot Chips on August 24 a processor whose cores execute Arm code and IBM Z code alike.
The chip, co-designed with Arm for future IBM Z and LinuxONE servers, drops the usual approach of bolting an Arm subsystem onto a Z machine. Instead each core juggles the two environments at once, with hardware deciding which gets priority, so customers can run both software stacks on one box.
It is built on a 2nm process with 11 high-performance cores clocked past 5.7GHz and AI inference accelerators that score transactions for fraud while they are in flight.
IBM’s Christian Jacobi, chief architect for IBM Z, cast the chip as an answer for enterprises that want AI-era workloads without giving up their core systems.
Mohamed Awad, who leads cloud AI at Arm, sees the industry converging on Arm as AI scales, pointing to the 22 million-plus developers in its ecosystem.
For IBM the payoff is simpler: Arm-native applications land on flagship mainframes and LinuxONE without rework, widening the software menu for cloud-native and AI workloads.
