Spyre Accelerator for Power: IBM i Shop Prerequisites
Spyre is a PCIe-attached inference accelerator for supported Power 11 configurations, not a native IBM i feature. Check the expansion drawer, Linux runtime, Red Hat stack, and resource minimums.
IBM Spyre Accelerator for Power is a PCIe-attached inference card offered for supported Power 11 systems. IBM documents an ENZ0 expansion drawer and a Red Hat inference stack. For an IBM i shop, the key distinction is that the documented accelerator environment runs in a supported Linux and Red Hat setup, not natively as an IBM i feature. Your IBM i workload may supply data, but it is not itself the Spyre runtime.
Requirements to check before a quote
Sources: IBM Spyre introduction and IBM hardware and software requirements. These are current-documentation checks, not a substitute for a supported-machine quote.
When Spyre is not the next step
If the first use case is IBM Bob for RPG code assistance, an external watsonx application, or a small local model that a supported Power CPU runtime can handle, the Spyre drawer may not answer the actual need. Pilot the workload, confirm where it runs, and size the system from measured throughput. On-chip MMA and Spyre solve different capacity problems.
Start with AI workload placement and the infrastructure readiness checklist. For software AI tools, use AI.as400Software.com. For Power 11 generation and model detail, see the Power 11 specialist.