IBM Artificial Intelligence Unit (AIU)
The research chip that became the card you can order. The AIU is why the Spyre Accelerator has 32 cores and why those cores look so much like the AI engine already inside an IBM Z processor.
The Artificial Intelligence Unit is the prototype. It is what IBM Research built when it wanted to find out whether a whole chip dedicated to running AI models could earn its place in an enterprise server.
The answer turned out to be yes, and the proof is that you can now order the result. The Spyre Accelerator is a tuned version of the AIU design. Same 32 cores. Same idea of giving each core its own small block of fast scratchpad memory instead of making all of them queue for one shared pool.
Why the family tree matters to a buyer
Because it tells you what the card is good at. The AIU cores were closely related to the AI engine IBM had already put inside the Telum processor for mainframe fraud detection. That heritage is about scoring transactions in the time it takes one to clear, not about training enormous models from scratch.
So when you evaluate a Spyre card, evaluate it as a very fast answering machine sitting next to your data. That is the job its family was bred for.
The AIU is not a product. Treat it as background that explains the Spyre Accelerator's design choices, not as something to specify or price.
Related
IBM Spyre Accelerator for Power
A 75 watt PCIe card that adds more than 300 TOPS of AI inference to a Power 11 server. It is the only IBM AI chip an AS/400 shop can actually order as an add-on, and it comes with a short list of things it refuses to work without.
IBM NorthPole
A research chip with no separate memory at all. NorthPole claims 25 times the energy efficiency of a conventional design by refusing to move data, and IBM has already run a Granite model on it.
IBM Telum II Processor
The mainframe processor that proved IBM's whole AI strategy. Telum II does not go in a Power server, but the AI engine inside it is the reason the Spyre card exists, and the reason IBM talks about inference in milliseconds rather than seconds.
Power 11 Matrix Math Acceleration
The AI acceleration most AS/400 shops already own and have never switched on. It is inside every core of the processor, it costs nothing extra, and it is the cheapest honest answer to a first AI pilot.