IBM has unveiled the IBM z17 mainframe designed with built-in AI capabilities to support advanced AI workloads. IBM did not reveal the price of z17 — powered by the new Telum II processor.
z17 delivers 50 percent more AI inference operations per day compared to its predecessor, the z16, and supports over 250 AI use cases across industries — from loan risk mitigation and medical image analysis to chatbot management and retail crime prevention.
Developed over five years and backed by more than 300 patent filings, the z17 reflects input from over 100 clients and collaboration with IBM Research and Software teams. It introduces multi-model AI, enhanced security features, and new AI-driven tools to improve system usability and management. With this release, IBM aims to redefine AI at scale and enable enterprises to process 100 percent of their transactions in real-time.
The IBM z17 features advanced AI inferencing capabilities powered by the second-generation on-chip AI accelerator built into the Telum II processor. With increased frequency, compute power, and a 40 percent larger cache, this accelerator enables the system to perform over 450 billion AI inferencing operations per day with a rapid one millisecond response time.
To further enhance AI performance, IBM plans to introduce the IBM Spyre Accelerator — a PCIe card expected in Q4 2025. This accelerator will complement the Telum II processor by supporting multi-model AI methods and bringing generative AI capabilities to the mainframe, such as running AI assistants using enterprise data.
IBM z17 uses AI to improve developer and IT operations efficiency. It includes tools like IBM watsonx Code Assistant for Z and watsonx Assistant for Z, now integrated with Z Operations Unite. This integration enables AI-powered, chat-based incident detection and resolution, utilizing real-time system data to streamline operations and enhance user experience.
The z/OS 3.2, expected in the third quarter of 2025, will support hardware-accelerated AI, modern data access methods including NoSQL databases, and hybrid cloud data processing, allowing AI applications to extract deeper insights from enterprise data.
IBM introduced Z Operations Unite, which will unify performance metrics and logs using OpenTelemetry to streamline operations through AI-powered anomaly detection and incident resolution. When used with IBM Concert, it enhances the intelligent correlation of operational data across the enterprise.
Additionally, the IBM Spyre Accelerator, expected in late 2025, will provide expanded AI compute capabilities via PCIe, enabling generative AI workloads and AI assistants to run directly on the z17. This integration enhances productivity while maintaining data security and operational efficiency.
IBM z17 continues the platform’s legacy of security and resiliency, enhanced through new AI-driven capabilities that address the cyber threats. Among these advancements is the integration of secrets management via IBM Vault, powered by technology from HashiCorp, which standardizes identity-based access control for sensitive data like certificates, keys, and tokens across hybrid cloud environments. IBM is also introducing AI-powered tools to discover and classify sensitive data using Telum II and natural language processing, as well as IBM Threat Detection for z/OS to identify anomalies that could signal cyber-attacks.
IBM offers AI-enabled lifecycle support services through IBM Technology Lifecycle Services. These services help optimize system performance, minimize risk, and accelerate incident resolution using IBM watsonx.
IBM z17 is integrated with the latest generation of IBM Storage DS8000, delivering high-performance, secure, and flexible storage for mission-critical workloads.
Baburajan Kizhakedath
