IBM to Launch z17, First Mainframe for AI Workloads

IBM is launching the z17, a next-generation mainframe system that integrates AI capabilities across hardware, software, and operations. The model features the new IBM Telum II processor and is designed to meet the real-time processing demands of enterprise-scale AI workloads.

"With z17, we are bringing AI to the core of the enterprise with the software, processing power, and storage to operationalize AI quickly," writes Ross Mauri, General Manager, IBM Z and LinuxONE, in a press release. Mauri emphasizes that z17 architecture aims to convert large volumes of unstructured enterprise data into operational value securely and cost-effectively, directly where the data resides.

The development of the IBM z17 addresses several critical challenges in implementing AI in corporate environments. As AI expands from specific applications into integrated enterprise systems. Demand is increasing for platforms capable of handling large data volumes, real-time transactions, and strict security requirements.

“Traditional infrastructures, designed for earlier computing models, face technical limitations when executing complex generative AI models, predictive analytics, or anomaly detection in mission-critical contexts,” says Mauricio Torres, General Director and President, IBM Mexico. “The mainframe architecture — recognized for reliability, scalability, and security — has historically been adopted by sectors such as financial services, healthcare, insurance, and retail.”

With z17, IBM aims to position the mainframe not only as a transactional platform but also as the computational core for complex AI workloads. Its objective is to enable real-time analysis of 100% of enterprise transactions, according to data provided by IBM.

The system is the result of a five-year development process, with over 300 patent filings and direct collaboration with more than 100 clients, along with IBM Research and IBM Software teams.

Technical Details and Capabilities of the IBM z17

The IBM z17 is equipped with the IBM Telum II processor, which includes a second-generation on-chip AI accelerator. This component offers substantial improvements over its predecessor, including a 40% increase in cache memory and enhanced computing capabilities. The system can process over 450 billion inference operations per day, with 1ms response times and a 50% performance improvement over the z16. These features support over 250 use cases defined by IBM, including credit risk mitigation, medical image analysis, fraud prevention, and optimization of customer service through conversational AI.

In addition to the Telum II, IBM has announced the IBM Spyre Accelerator PCIe card, expected to be available in the fourth quarter of 2025. This unit will provide additional AI processing power and allow generative model-based execution assistants to run directly on data stored in the mainframe, eliminating the need to transfer it to external environments.

Among the enabled assistants are IBM watsonx Code Assistant for Z and IBM watsonx Assistant for Z, the latter integrated with Z Operations Unite. These solutions aim to increase development and operations team efficiency through automation and conversational assistance based on real-time system data.

Hybrid Operations, Software Support, and Security

IBM also introduced enhancements related to integrating the z17 into hybrid cloud environments. The system is designed to support modern data access methods, NoSQL databases, and processing across multiple clouds. In this context, the company announced z/OS 3.2, the upcoming version of the operating system for IBM Z, scheduled for release in the third quarter of 2025. This version will provide full support for hardware-accelerated AI, AI-based operational insights, and connectivity with modern services.

For operational management, IBM announced the general availability of IBM Z Operations Unite starting in May 2025. This solution consolidates key performance metrics and logs across the platform using the OpenTelemetry format, enabling AI-based incident detection and resolution. Used in combination with IBM Concert, it is expected to significantly reduce diagnostic and resolution times.

In terms of security capabilities, IBM z17 extends existing functionality with integrated HashiCorp technologies for the protection and management of credentials, certificates, and keys. It also introduces sensitive data classification capabilities using AI and natural language processing, supported by the Telum II processor. Additionally, IBM Threat Detection for z/OS can identify anomalies that could indicate cyberattacks by detecting irregular patterns.

The IBM z17 can be integrated with the tenth generation of the IBM Storage DS8000 system. This unit is designed for critical workloads, with a modular architecture and adaptability to emerging technologies to facilitate the monetization of enterprise data.

IBM Technology Lifecycle Services will provide support for z17 through AI-based automated processes using IBM watsonx to optimize incident resolution times.

In Mexico, the IBM z17 will be commercially available starting June 18, 2025. The IBM Spyre Accelerator card will be available in the fourth quarter of 2025.