IBM Unveils AI-Optimized Mainframe for Modern Era

# IBM

IBM has just unveiled the IBM z17, the latest addition to its iconic mainframe lineup, and it's packing some serious AI muscle. Announced on Monday, this fully encrypted powerhouse is driven by the new IBM Telum II processor and is tailored for over 250 AI applications. From AI agents to generative AI, the z17 is set to supercharge your AI endeavors. You might think mainframes are a thing of the past, but think again. A whopping 71% of Fortune 500 companies still rely on them, and the market was valued at a cool $5.3 billion in 2024, according to Market Research Future. Clearly, mainframes are still very much in the game. The z17 isn't just about keeping up; it's about setting the pace. It can handle a staggering 450 billion inference operations daily, marking a 50% leap from its predecessor, the IBM z16, which hit the market back in 2022 with the original Tellum processor. Plus, the z17 plays nice with other hardware, software, and open-source tools, making it a versatile beast. Tina Tarquinio, IBM Z's VP of product management and design, shared with TechCrunch that this upgrade has been a five-year journey. And get this—it started well before the AI craze kicked off with OpenAI's ChatGPT in November 2022. IBM poured over 2,000 research hours into gathering feedback from more than 100 customers during the development of the z17. Tarquinio finds it fascinating that their early insights aligned so well with the market's eventual direction. "It's been wild to know we were working on an AI accelerator and then witness the industry's AI transformation, especially in late 2022," Tarquinio told TechCrunch. "It's been thrilling. The biggest thing is we don't know what we don't know about the future, so the possibilities with AI are truly endless." The z17 is designed to evolve with the AI landscape. It'll launch with support for 48 IBM Spyre AI accelerator chips, with plans to double that to 96 within a year. "We're building in headroom and AI agility," Tarquinio explained. "As new models emerge, we're ensuring we have the space for bigger, more complex models that might need more local memory to interact. We're ready for the shifts in approach because new models will keep coming." When asked about her favorite feature of the z17, Tarquinio playfully compared it to choosing a favorite child. But seriously, she highlighted its energy efficiency. "On-chip, we're boosting AI acceleration by seven and a half times, but that's with five and a half times less energy than you'd need for multi-model processing on other accelerators or platforms," she said. Mark your calendars: the IBM z17 mainframes will be hitting the market on June 8. Get ready to take your AI game to the next level with this powerhouse.

Related article

Comments (26)