IBM’s next-gen mainframe chip is the first to run Arm and Z workloads on the same cores

IBM announced at the Hot Chips conference a new mainframe processor whose cores can natively execute both IBM’s own Z instruction set and Arm’s, switching between the two in nanoseconds. The chip, destined for the next generation of IBM Z and LinuxONE systems, is the first dual-architecture mainframe processor ever built. It lets enterprises run Arm-native Linux software, including AI frameworks, alongside z/OS transaction-processing workloads used by banks, insurers, and governments. Tina Tarquinio, chief product officer for IBM Z and LinuxONE, called it one of the most powerful commercially available dual-architecture processors.

The chip has 11 cores, and each core can dynamically switch between Arm software mode and traditional Z software mode, according to Christian Jacobi, IBM Fellow and chief technology officer of IBM Systems Development. The switching mechanism relies on the open-source KVM hypervisor, which dispatches Arm64 Linux virtual machines and Linux on Z virtual machines onto physical cores, flipping the core into the corresponding mode. Jacobi said the switch takes about a nanosecond and has effectively no performance impact, since virtual images run for milliseconds between switches. Traditional z/OS workloads run in a separate partition on the same chip, outside KVM.

IBM chose not to add standalone Arm cores to the side of the processor, a simpler heterogeneous design used by other chipmakers. Jacobi said that approach would not address customer needs, and that deep integration into the entire system was required to deliver the same quality of service clients expect. The chip is built on a 2-nanometer process node and runs its 11 cores at a base frequency above 5.7 GHz, with on-chip AI inference accelerators for in-transaction fraud detection, a dedicated data processing unit for I/O acceleration, and a large cache architecture. Full systems scale to hundreds of cores and tens of terabytes of memory.

The strategic rationale is software, not hardware. IBM’s s390x architecture handles a large share of the world’s mission-critical transactions, but enterprise software such as monitoring tools, security agents, cloud-native middleware, and AI stacks like PyTorch and ONNX Runtime were built for x86 and increasingly for Arm. Arm estimates that close to half of compute shipped to major hyperscalers in 2025 was Arm-based, driven by AWS Graviton, Google Axion, and Microsoft’s Arm silicon, and Arm counts more than 22 million developers worldwide. Tarquinio said IBM’s ecosystem team could never port all those applications, so the company took a step-function approach from a technology standpoint.

Customers did not ask for a dual-architecture chip specifically, Tarquinio said, but requested help getting surround workloads to market faster. The compatibility promise is ambitious: Arm Linux binaries should run unmodified, Jacobi said, with 100% binary compatibility. Arm defines the instruction set architecture and supplies validation tooling to ensure IBM’s implementation behaves identically to other Arm chips, while IBM designs and builds the silicon in-house.

IBM also previewed the next generation of its Spyre AI accelerator at Hot Chips. The new Spyre is a higher-performance chip capable of running large language models for agentic workflows, Jacobi said, including AI-ops and business workflows like document understanding and insurance adjudication. It will ship with high-bandwidth memory.

Mohamed Awad, Arm’s executive vice president for cloud AI, said that as AI scales, more computing is converging on Arm, and bringing Arm compute to mission-critical enterprise infrastructure gives organizations greater choice in deploying AI. The announcement follows McKinsey data showing that while 88% of organizations use AI in at least one business function, nearly two-thirds have not scaled it across the enterprise, with the most value coming from companies that redesign core workflows. For regulated industries with systems of record on IBM Z, running AI where transactions happen is the direct application of this announcement.

More hardware news from TechManNews.