IBM is revolutionizing the mainframe landscape with its latest offering that combines traditional enterprise computing power with advanced artificial intelligence capabilities. The company's newest system represents a significant evolution in how businesses can process mission-critical workloads while leveraging AI directly on their core infrastructure.
 |
|---|
| Advanced circuit boards highlighting the cutting-edge technology of IBM's latest system |
The Next Generation Mainframe
IBM has introduced the z17, its newest mainframe system designed for mission-critical business transactions with enhanced security capabilities powered by artificial intelligence. This latest iteration marks a significant advancement in IBM's mainframe technology, positioning these systems not as relics of the past but as cutting-edge platforms for future innovation. The z17 maintains the fundamental qualities that have defined mainframes for decades—reliability, security, and performance—while incorporating new AI capabilities that address emerging business needs in an era of digital transformation.
Telum II: The Heart of Z17
At the core of the z17 is IBM's Telum II processor, featuring eight advanced cores operating at 5.5 GHz with enhanced branch prediction, store writeback, and address translation capabilities. The processor boasts 36MB of L2 cache, representing a 40% increase compared to its predecessor. It also supports virtual L3 and L4 cache levels, expanding available cache to 360MB and 2.88GB, respectively. Manufactured using Samsung's 5HPP fabrication process, the Telum II contains an impressive 43 billion transistors and integrates a data processing unit (DPU) to accelerate transactional workloads, improving overall system responsiveness.
AI Performance Leap
The z17 delivers 50% more AI inference capacity than its predecessor, the z16. A central element of the Telum II processor is its upgraded AI unit, which delivers four times the compute capability of the previous generation, reaching 24 trillion operations per second with INT8 data precision. While this may not seem impressive compared to dedicated AI accelerators, the NPU is specifically designed for mission-critical time-sensitive applications that support ensemble AI methods, combining traditional machine learning with large language models to detect suspicious activities and fraud attempts in real-time.
Spyre AI Accelerator
Understanding that some workloads will require additional AI performance, IBM has unveiled the Spyre AI accelerator card with a PCIe interface. This 26-billion transistor processor features 32 AI cores and an architecture similar to the AI accelerator found in Telum II. The Spyre Accelerator enables native support for large language models and other advanced AI applications directly on the mainframe, eliminating reliance on external GPUs or cloud services. This add-on component can dynamically expand the AI capabilities and performance of z17 systems, making it possible to run generative AI workloads directly on the mainframe.
Software Enhancements
To complement the z17 hardware, IBM will introduce z/OS 3.2 in Q3 2025, an updated version of its mainframe operating system. This AI-optimized OS is designed to work with hardware acceleration and supports NoSQL and hybrid cloud data integration. The system also includes IBM Vault, originally developed by HashiCorp (which IBM recently acquired), to handle credentials, keys, and tokens across hybrid environments. Additionally, operational management is enhanced through IBM Z Operations Unite, which uses OpenTelemetry to unify and streamline observability and incident response.
Enterprise AI Strategy
By keeping AI workloads on-premises, the z17 offers businesses greater control over sensitive data, reduced architectural complexity, and faster, more secure processing. This approach is particularly valuable in industries with strict regulatory requirements and high-security standards, such as banking, healthcare, and government. The z17 is also deeply integrated with IBM's watsonx platform, including tools like watsonx Code Assistant for Z and watsonx Assistant for Z, which support the modernization of legacy COBOL applications by simplifying code maintenance and accelerating refactoring efforts.
Mainframe Relevance in Modern IT
Despite perceptions of mainframes as outdated technology, they continue to serve as the backbone of industries where uptime, throughput, and security are non-negotiable. A recent IBM-sponsored survey found that 78% of global IT executives consider the mainframe central to their digital transformation strategies, while 88% consider application modernization a key priority. In today's environment—characterized by rising cloud costs, heightened privacy concerns, and increasing interest in AI-enabled decision-making—many enterprises are reconsidering the value of maintaining secure, on-premises infrastructure.
Availability and Positioning
The IBM z17 will be available starting June 18, 2025, with the Spyre Accelerator arriving later in the year during Q4 2025. While companies like Nvidia dominate the GPU market and cloud providers such as AWS, Microsoft Azure, and Google Cloud lead the AI services space, IBM is carving out a different path with the z17. By offering an on-premises, vertically integrated solution tailored for industries that demand regulatory compliance, data control, and uninterrupted operations, IBM has positioned itself uniquely in the evolving AI infrastructure landscape.