Glossary

IBM Spyre Accelerator

Spyre

A PCIe4 AI inferencing card IBM ships for IBM Z, LinuxONE and Power11 systems to offload generative and agentic AI workloads from the main processor. On Power11 it is driven from a Red Hat Enterprise Linux or Red Hat OpenShift AI partition, and as of September 2026 IBM has not documented a way to use it from an IBM i partition.

IBM introduced the Spyre Accelerator for commercial availability on October 7, 2025, reaching general availability October 28, 2025 for IBM z17 and LinuxONE 5, and early December 2025 for Power11. Each card carries 32 AI-optimized cores and 25.6 billion transistors on a 5nm process within a 75-watt envelope; a Power11 server holds up to 16 cards in an ENZ0 PCIe4 expansion drawer, versus up to 48 on IBM Z and LinuxONE. It is a separate, dedicated inferencing card, distinct from the Matrix Math Acceleration built into every Power11 core. IBM's Spyre documentation lists Red Hat AI Inference Server on RHEL 9.6, 9.8 or 10.2, or Red Hat OpenShift AI, as the required software stack, so a Power11 owner accesses Spyre from a Linux or OpenShift partition. Neither IBM's documentation nor its October 2025 announcement mentions IBM i, and as of September 2026 IBM has not published a route for an IBM i partition to use an installed Spyre card directly.

Related Terms