PCIe AI Accelerator Card

IBM Spyre Accelerator

IBM's PCIe-based AI accelerator card, built for on-premises generative and traditional AI inference, with confirmed Power11 compatibility arriving December 2025.

The Spyre Accelerator is IBM's PCIe-form-factor AI accelerator, designed to bring on-premises AI inference -- both generative and traditional model workloads -- directly into enterprise transaction processing. It shares its AI silicon lineage with the Telum II on-chip accelerator used in IBM's z17 mainframe.

IBM confirmed Spyre Accelerator compatibility with Power11 systems arriving in December 2025. More significantly for anyone tracking Power12, IBM Power Processor Chief Architect Bill Starke told IT Jungle that Spyre "will be a piece of the Power11 portfolio and beyond," alongside continued evolution of IBM's WatsonX models. That is a direct, attributed IBM statement -- not analyst speculation -- that AI acceleration on Power hardware is a multi-generation commitment, not a one-off Power11 feature.

What this does and doesn't confirm

It confirms IBM's AI acceleration strategy continues past Power11. It does not confirm any Power12 model number, core count, TOPS figure, or release date -- IBM has not published any of those.

Frequently Asked

Will Power 12 have on-chip AI acceleration?

This is the most directly confirmed forward-looking AI signal available: IBM Power Processor Chief Architect Bill Starke has stated the Spyre AI Accelerator "will be a piece of the Power11 portfolio and beyond," alongside continued evolution of IBM's WatsonX models. The specific form that takes for the next generation -- expanded on-chip acceleration, a next-generation Spyre card, or otherwise -- has not been disclosed.