IBM is releasing the most recent model of its mainframe {hardware} that features new updates meant to speed up AI adoption.
The {hardware} and consulting firm on Monday introduced IBM z17, the most recent model of its mainframe pc {hardware}. This absolutely encrypted mainframe is powered by an IBM Telum II processor and is designed for greater than 250 AI use circumstances, the corporate says, together with AI brokers and generative AI.
Mainframes may look like outdated hat, however they’re utilized by 71% of Fortune 500 corporations in the present day, in line with one supply. In 2024, the mainframe market was price an estimated $5.3 billion, per consulting agency Market Analysis Future.
The z17 can course of 450 billion inference operations in a day, a 50% improve over its predecessor, the IBM z16, which was launched in 2022 and ran on the corporate’s authentic Tellum processor. The system is designed to have the ability to absolutely combine with different {hardware}, software program, and open-source instruments.
Tina Tarquinio, VP of product administration and design for IBM Z, advised TechCrunch that this mainframe improve has been within the works for 5 years — properly earlier than the present AI frenzy that began with the discharge of OpenAI’s ChatGPT in November 2022.
IBM spent greater than 2,000 analysis hours getting suggestions from over 100 clients because it constructed the z17, Tarquinio mentioned. She thinks it’s attention-grabbing to see that, now, 5 years later, the suggestions they bought aligned with the place the market ended up heading.
“It has been wild figuring out that we’re introducing an AI accelerator, after which seeing, particularly within the later half of 2022, the entire adjustments within the business concerning AI,” Tarquinio advised TechCrunch. “It’s been actually thrilling. I feel the most important level has been [that] we don’t know what we don’t find out about what’s coming, proper? So the chances are actually limitless when it comes to what AI might help us do.”
The z17 is about as much as adapt and accommodate the place the AI market heads, Tarquinio mentioned. The mainframe will assist 48 IBM Spyre AI accelerator chips upon launch, with the plan to deliver that quantity as much as 96 inside 12 months.
“We’re purposely constructing in headroom,” Tarquinio mentioned. “We’re purposely constructing in AI agility. In order new fashions are launched, [we’re] ensuring that we’ve constructed within the headroom for larger, bigger fashions — fashions that perhaps want extra native reminiscence to speak to one another. We’ve in-built that as a result of we all know it’s actually the strategy that can change, proper? The brand new fashions will come and go.”
Tarquinio mentioned that one of many highlights of this newest {hardware} — though she joked it was like being requested to choose her favourite baby — is that the z17 is extra energy-efficient than its predecessor and supposedly opponents, too.
“On-chip, we’re rising the AI acceleration by seven and a half instances, however that’s 5 and a half instances much less power than you would wish to do, like, multi-model on one other kind of accelerator or platform within the business,” Tarquinio mentioned.
The z17 mainframes will change into usually obtainable on June 8.
