IBM unveiled at Hot Chips the design of their first dual-architecture processor for future IBM Z and LinuxONE systems, capable of running Arm instructions natively alongside those of IBM’s enterprise platforms. The chip, still in development, will be fabricated using 2-nanometer technology, will feature 11 cores operating above 5.7 GHz, and will include AI acceleration and input/output processing directly on the silicon.
The key points of the IBM-Arm processor in 30 seconds
- IBM is developing a processor capable of executing Arm and IBM Z or LinuxONE instructions on the same cores, without physically separating the two architectures.
- It will be manufactured using 2 nanometer technology, with 11 high-performance cores running above 5.7 GHz.
- It will enable combining Arm Linux environments with z/OS and Linux on IBM Z.
- It will incorporate accelerators for AI inference and a dedicated unit for input/output operations.
- This is the first tangible result of the collaboration announced by IBM and Arm in April 2026.
The announcement is especially significant because IBM isn’t simply planning to add auxiliary Arm processors to its mainframes. The architecture described by the company aims for each core to understand and execute two different sets of instructions natively and simultaneously.
The goal is to bring applications developed for the broad Arm ecosystem to traditional systems associated with transaction processing, large databases, and critical enterprise workloads.
Arm directly enters IBM’s mainframe core
IBM Z uses an architecture with decades of evolution. While modern mainframes look quite different from their predecessors, they maintain a hardware and software environment designed around very specific requirements for availability, security, and transactional processing.
This new approach introduces a significant change.
IBM explains that the future processor will not use a separate group of Arm cores and independent IBM cores. Instead, each core will be designed to execute instructions for both Arm and IBM Z or LinuxONE.
This would allow a single system to run native Arm Linux applications alongside z/OS and Linux on IBM Z concurrently.
Although the difference may seem technical, it has practical implications. Currently, Arm boasts a vast catalog of tools, frameworks, containers, and applications created for cloud servers, devices, and AI systems.
According to IBM, the Arm ecosystem exceeds 22 million developers worldwide.
The company aims to bring that software to its enterprise platforms without requiring the abandonment of features that distinguish IBM Z and LinuxONE, such as hardware fault detection and recovery, encryption, secure key management, and high availability capabilities.
This initiative stems from a strategic collaboration announced by IBM and Arm on April 2, 2026. At that time, both companies indicated they would explore dual-architecture hardware and virtualization technologies enabling the deployment of Arm applications on IBM’s enterprise systems.
Four months later, IBM has revealed the first concrete design derived from that agreement.
2 nanometers, over 5.7 GHz, and AI acceleration
The initial specifications also give an idea of the kind of processor IBM is developing.
The chip will be built using a 2-nanometer process technology and will feature 11 high-performance cores capable of operating above 5.7 GHz.
IBM has not yet detailed all its features nor announced a specific release date for the first commercial systems utilizing it.
It has confirmed, however, that it will include AI inference accelerators, continuing a technological line already underway with processors like Telum II.
One of the explicit applications mentioned by IBM is fraud detection during transactions. The idea is to perform certain AI operations directly where the data is being processed, avoiding the need to transfer data to external infrastructure for analysis.
The processor will also incorporate a Data Processing Unit (DPU) specifically designed to accelerate input/output operations, along with a large-capacity cache architecture aimed at intensive enterprise workloads.
These features should be considered within the context of much larger systems. IBM notes that its Z and LinuxONE platforms can scale to hundreds of cores and tens of terabytes of memory.
Therefore, the objective is not to compete directly with conventional Arm processors used in individual servers. Instead, IBM seeks to integrate the Arm ecosystem into machines designed to consolidate large amounts of enterprise processing.
Arm’s expansion into a new territory of x86 and proprietary architectures
The move also reflects a broader trend in the server market.
Arm has been expanding its presence beyond smartphones and embedded devices for years. Amazon Web Services develops its Graviton processors for cloud infrastructure, NVIDIA has designed Vera for its future AI platforms, and other manufacturers are increasing investments around this architecture.
IBM is taking a different approach.
Rather than replacing its traditional architecture with Arm, it aims for both to coexist within the same processor.
This would allow retaining enterprise applications developed over decades while integrating modern workloads originally built for Arm.
The strategy could be especially promising in AI and cloud-native applications—two fields where software compatibility considerations heavily influence architecture choices.
It could also reduce one of the common barriers of specialized systems: the need to adapt or recompile applications for a different platform.
However, IBM still needs to clarify in more detail how this compatibility will work in real-world scenarios, which Arm operating systems will be supported, how resource management between architectures will be handled, and what the performance impact of simultaneous execution will be.
The company emphasizes that these are future plans subject to change, and the features announced should not be considered final specifications of a commercial product already available.
What was introduced at Hot Chips is primarily a fairly concrete statement about the direction future IBM Z and LinuxONE generations will take.
For years, mainframe modernization has mainly involved integrating Linux, containers, and cloud tools into a historic platform. With this processor, IBM aims to take a further step: enabling some of the software designed for Arm to reach the mainframe without Arm becoming an external platform.
Frequently Asked Questions
Will the new IBM processor have separate Arm cores?
No. According to IBM, each core will be designed to natively execute both Arm and IBM Z or LinuxONE instructions, rather than dividing the processor into two separate core groups.
What are the specifications of the processor?
IBM revealed a design of 2 nanometers with 11 high-performance cores over 5.7 GHz, including AI inference accelerators, an integrated DPU for input/output, and a broad cache architecture.
Will it run Arm Linux applications?
That is one of its main goals. IBM wants to enable future systems to run native Linux Arm environments concurrently with z/OS and Linux on IBM Z.
When will it arrive on IBM Z and LinuxONE servers?
IBM has not announced a concrete commercial release date yet. The company describes the processor as a technology in development aimed at future generations of IBM Z and LinuxONE.
