Logical Partition (LPAR)
Splits one IBM Power server into separate virtual servers, each holding its own processors, memory and I/O devices, so IBM i, AIX and Linux can run side by side on the same machine. That split is how most IBM i shops add AI today: the model runs in a Linux partition next door and IBM i calls it over a web request.
What a logical partition does
IBM says partitioning is the ability to make a server run as if it were two or more independent servers. Picture one office building with four floors. Each floor has its own door, its own lock and its own thermostat. Nobody on the second floor can walk into the third. A logical partition is a floor. The building is the Power server, and the floors are handed out in processors, memory and I/O devices.
- Each partition runs its own operating system, and it boots and shuts down on its own.
- IBM names four things that run in a partition: AIX, IBM i, Linux, and the Virtual I/O Server.
- Resources are handed out as processors, memory and I/O devices.
- The Hardware Management Console, or HMC, is the console that creates and manages the partitions.
- A partition that shares processor capacity instead of owning whole cores is called a shared processor partition, or a micro-partition.
How small can a partition be?
Smallest slice of one processor core a shared partition can use
IBM calls 0.10 about one tenth of the capacity of a single processor, and 0.05 about a twentieth. In between, IBM says you can set processing units down to a hundredth of a unit. A smaller number is the better one here, because it lets a small job take a thin slice instead of a whole core.
The size is not locked once the partition starts. IBM says you can add, remove and move physical memory to and from running partitions that use dedicated memory with the HMC, without having to shut down the partitions. Processors move the same way. So a Linux partition doing AI work can borrow cores for a busy hour and give them back, while the IBM i partition beside it keeps running.
How does AI reach IBM i through a partition?
Four steps in IBM's fraud detection write-up, dated October 24, 2025
- Build the Linux partitionNine cores of a Power E1180 went to a Linux partition. The IBM i partition beside it used a single core. IBM placed both partitions on the same socket, meaning the same physical processor chip.
- Put the model behind a web addressAn LSTM model ran behind a Flask REST API inside a podman container. LSTM is short for long short-term memory, a model type that reads a sequence in order instead of one item at a time.
- Let IBM i ask the questionA trigger fired when a transaction row was inserted. The trigger built a JSON document holding the new transaction plus the previous six for that user and card.
- Read the answer backThe trigger sent the JSON document to the REST endpoint on the Linux partition, and the model answered whether the current transaction looked fraudulent.
IBM reports a model load time averaging 90 ms, about 0.11 ms to pull the JSON data, and around 2 percent processor use on the Linux partition. IBM's own disclaimer dates the test to July 1, 2025. These are IBM's measurements on IBM's setup, not a promise for yours.
Is there a tool that builds the AI partition for you?
Yes. IBM publishes an open project called Power Inference Microservices, or PIM, on GitHub as IBM/project-pim. It creates new Linux partitions on a Power server through the HMC and boots them from a bootable container, which is a container image that starts as a whole operating system instead of running inside one. IBM ships pre-built images for vLLM, an inference engine, and describes the goal as spinning up an AI environment with very little user intervention, next to workloads already running on the server.
- Needs a Hardware Management Console to create and manage the new partition.
- Uses bootc, the bootable container technology, to install the Linux partition.
- Supports IBM Power 10 and Power 11 servers, with minimum levels of HMC 1061 for Power 10 and HMC 1110 for Power 11.
- Lists minimum IBM i levels of 7.5 for Power 10 and 7.6 for Power 11.
- Also sets minimum partition firmware and Virtual I/O Server levels.
- Covers updates, rollback and other lifecycle steps, not only the first install.
PIM is an IBM GitHub project, not a priced IBM product with a support contract. Read its own minimum levels before planning around it, because they move as new Power firmware and IBM i Technology Refresh levels ship.
What limits an IBM i partition?
The server model sets the ceiling, not the partition itself. The small end of the Power 11 line shows this clearly.
What IBM's Power S1112 overview states about partitions on that model
| Item | What IBM states |
|---|---|
| Software tier | P05, and the S1112 is the only Power 11 server offered at that tier |
| IBM i processor ceiling | All IBM i partitions together are limited to 4 cores |
| IBM i memory ceiling | All IBM i partitions together are limited to 64 GB |
| IBM i releases listed | IBM i 7.6 TR 2, IBM i 7.5 TR 8, IBM i 7.4 TR 12 |
| Linux releases listed | RHEL 9.6, RHEL 9.8, RHEL 10.2, SLES 15 SP 7, SLES 16 |
| Spyre accelerator | Not supported on the S1112 |
These figures are for the Power S1112 only. Larger Power 11 servers carry different ceilings, and IBM's overview notes that scale-out servers above the S1112 support an I/O expansion drawer for up to 8 Spyre adapters. Matrix Multiply Assist, the AI math built into the Power chip, is named as present either way.
Related Terms
IBM Power 11
IBM's current generation of Power processor and server hardware, announced July 8, 2025 and generally available July 25, 2025, carrying forward Power10's on-chip Matrix Math Acceleration and becoming the first Power generation to support the IBM Spyre Accelerator card.
IBM Power 10
IBM's Power processor generation that preceded Power11, announced August 17, 2020 and shipping from September 2021, the first Power chip to embed Matrix Math Acceleration into every core for on-chip AI inferencing without a separate GPU. IBM began winding down new Power10 sales in February 2026.
IBM Power S1112
IBM's entry-level, single-socket Power11 server, announced July 15, 2026 and generally available July 24, 2026, built for edge and branch AI inferencing with on-chip Matrix Math Acceleration. It is the first Power11 machine available in IBM i's low-cost P05 software tier.
IBM i
IBM's operating system for Power servers running traditional AS/400-lineage business applications, with Db2 for i built directly into the OS.
IBM AIX
AIX is IBM's UNIX operating system for IBM Power hardware, the sibling of IBM i on the same servers and the source of the AIX runtime that IBM i ships as PASE for i. On Power 11 it needs AIX 7300-03-01 or higher to run in Power11 compatibility mode, and its AI packages arrive as RPMs from the AIX Toolbox for Open Source Software.
Db2 for i
The relational database management system built into every IBM i operating system, tightly integrated with IBM i rather than a separately installed database product.
vLLM
An open source server that keeps a large language model loaded and answers requests over an OpenAI style API, with PowerPC CPUs on its supported hardware list. That lets an IBM i shop run a model in a Linux partition on the same IBM Power server as its production workload, and Red Hat ships a generally available vLLM container image for ppc64le in Red Hat AI Inference 3.5.0.
ppc64le
The name Linux, Red Hat packages, container images and Python wheels use for a 64-bit IBM Power processor running in little endian byte order, and the word an IBM i shop has to match when it installs AI software on Power hardware. IBM i itself is not ppc64le, so a ppc64le build runs in a Linux partition on the same Power server, never inside IBM i or PASE.
Red Hat AI Inference Server
A supported, packaged build of the open source vLLM server, so an IBM Power shop can serve its own large language models on hardware it already owns. Red Hat announced it on May 20, 2025 and IBM posted general availability on IBM Power on December 11, 2025, where it runs in a Linux partition rather than inside IBM i.
Matrix Math Acceleration (MMA)
An on-chip acceleration feature in IBM Power10 and Power11 processors that speeds up the matrix math operations used in AI inferencing, without requiring a separate GPU.
IBM Spyre Accelerator
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 Power Virtual Server (PowerVS)
IBM Power Virtual Server, or PowerVS, lets an IBM i shop rent IBM Power server capacity inside IBM's data centers instead of buying a box, and run the same AIX, IBM i and Linux partitions there. IBM describes it as a family of configurable, multitenant, virtual IBM Power servers with access to IBM Cloud services, web, DevOps and IBM Db2, which is how an IBM i partition ends up on the same cloud network as IBM Cloud AI services such as watsonx.ai.