IBM unveiled its latest mainframe hardware, the IBM z17, on Monday. This fully encrypted mainframe is powered by the IBM Telum II processor and designed for over 250 AI use cases, including AI agents and generative AI applications.

While mainframes might be considered legacy technology by some, a source reveals that 71% of Fortune 500 companies still use them today. According to Market Research Future, the mainframe market is estimated to be worth $5.3 billion by 2024.

The z17 can process 450 billion inferences daily, a 50% increase over its predecessor, the IBM z16, released in 2022. The new system is designed for seamless integration with other hardware, software, and open-source tools, providing a flexible enterprise computing solution.

Internet Big Data 2

Image Source Note: Image generated by AI, licensed through Midjourney.

In an interview with TechCrunch, Tina Tarquinio, IBM's VP of Product Management and Design for IBM Z, said the mainframe upgrade was five years in the making, long before the current AI boom sparked by OpenAI's November 2022 release of ChatGPT. IBM invested over 2,000 hours of research into building the z17, gathering feedback from over 100 clients. Tarquinio found it interesting how closely their five-year-old research direction aligned with the market's eventual trajectory.

Tarquinio stated: "When we learned we were going to have AI accelerators, especially in the latter half of 2022, all the changes in the industry around AI were incredibly exciting. I think the biggest question was, we didn't know what was coming next. So, in terms of what AI could help us do, the possibilities were truly limitless."

The z17's design anticipates the rapid evolution of the AI market. At launch, it will support 48 IBM Spyre AI accelerator chips, with plans to increase that to 96 within 12 months. "We intentionally built in headroom for increased AI agility," Tarquinio explained. "As new models come out, we ensure we have room for larger models—models that might need more on-chip memory to communicate with each other. We did that because we knew this was a truly evolving space. New models will continue to emerge and evolve."

Tarquinio highlighted the z17's energy efficiency as a key feature. "On-chip, we've improved AI acceleration by 7.5x, but the power consumption required is 5.5x less than what's needed for multi-model processing on other types of accelerators or platforms in the industry."

The IBM z17 mainframe will be generally available on June 8th.