IBM announced the launch of its new mainframe, the IBM z17, which integrates AI capabilities across hardware, software, and systems operations. This latest mainframe iteration, powered by the IBM Telum II processor, enhances the system's AI inferencing capabilities and supports more diverse workloads compared to its predecessor, the z16. According to the company, the IBM z17 can process 50% more AI inference operations per day than the z16.

The IBM z17 has been in development for five years, during which the company filed over 300 patents. Ross Mauri, general manager of IBM Z and LinuxONE, IBM, emphasized the importance of robust infrastructure for AI, stating, "The industry is quickly learning that AI will only be as valuable as the infrastructure it runs on." Mauri added that the z17 system is designed to operationalize AI swiftly while securely leveraging enterprise data.

In addition to enhancing AI processing, the z17 introduces multi-model AI capabilities, bolstered security features, and AI tools to improve system usability. A notable feature is the on-chip AI accelerator in the IBM Telum II processor, which increases frequency and computing capacity. Additionally, IBM plans to release a new AI compute accelerator, the Spyre Accelerator, via PCIe card by the fourth quarter of 2025.

The company also previewed the forthcoming IBM z/OS 3.2 operating system, which will include hardware-accelerated AI capabilities and support modern data access methods and hybrid cloud processing. The system aims to provide predictive business insights by accessing a broad set of enterprise data.

Security on the IBM z17 is also enhanced through capabilities such as IBM Vault for secrets management and AI-driven security solutions designed to detect potentially malicious anomalies. The platform's security and resiliency have been further strengthened with these developments.

IBM z17 will be available to clients on June 18, 2025. The release of the IBM Spyre Accelerator is scheduled for late 2025.