As AI accelerates across every layer of the enterprise, the mainframe is undergoing its most consequential redesign in decades — and hybrid AI infrastructure has found itself decidedly at the center of conversations.

IBM Corp. has spent years embedding AI directly into its hardware platforms, positioning IBM Z —its flagship line of mainframe computers — and its storage portfolio as the foundational layer for hybrid AI workloads that enterprises cannot afford to move to the public cloud. The combination of sovereign data requirements, agentic AI adoption and a decade of sustained Z growth has now placed IBM infrastructure in an unlikely position: one of the fastest-growing bets in enterprise technology, according to Ric Lewis (pictured), senior vice president of the Infrastructure Division at IBM.

“It couldn’t be more fun right now,” Lewis said. “The industry is as exciting as I’ve ever seen it in my entire career. Every boardroom is having a conversation around, ‘What are you doing on AI? What’s the plan? What’s the infrastructure that underlies that?’ The conversation has changed radically in just the last 18 months with agentic and how quickly that’s changing everything.”

Lewis spoke with theCUBE’s John Furrier and Dave Vellante at the Think 2026 event, during an exclusive broadcast on theCUBE, SiliconANGLE Media’s livestreaming studio. They discussed IBM’s hybrid AI infrastructure strategy, the decade-long growth of the Z platform and the strategic collaboration with Arm Holdings PLC. (* Disclosure below.)

IBM Z and hybrid AI infrastructure as the enterprise backbone

IBM’s Z platform has long been the backbone of the world’s most sensitive transaction processing, and the AI era is extending — not ending — that relevance. The company reported a 48% jump in mainframe revenue from its latest product line in mid-April, a figure that can be attributed directly to sustained client demand for AI-ready, on-premises infrastructure, according to Lewis. Z has grown 20 to 30% program to program for nearly a decade, driven by clients bringing new workloads onto the platform, he added.

“We have AI built on the chip. We have AI cards that we plug in the systems that we started shipping last year,” Lewis said. “All of that is resonating. The conversation we’re constantly having with clients is, ‘We’d like to bring more workloads to the platform, but we don’t want to have them just be the ones that run on Z OS.’ As long as it’s performant, we’d love to [put alternatives] on there. We’re making that a reality and designing it.”

That client feedback has been the direct origin of IBM’s collaboration with Arm to bring Arm-ecosystem applications natively into the next generation of Z systems — without emulation. The move is designed to let enterprises consolidate Arm-native workloads onto a single Z platform with the security, reliability and AI processing already built into IBM’s silicon, Lewis explained. The business case is already established by Linux app stacking on Z, he added.

“We’re integrating Arm technology directly, natively, into our chips,” Lewis said. “It’ll still run all the same Z apps the same way it does — it’ll just be able to execute Arm as well.”

The hybrid AI infrastructure imperative also shapes the broader landscape. AI has made hybrid computing an inevitability, because the data enterprises need is distributed across on-premises systems, neoclouds and hyperscalers, Lewis explained. IBM’s response has been to bring AI capability to the data — through built-in AI on Z and IBM Power and through storage systems such as IBM Storage Fusion that simplify data pipelines — rather than requiring enterprises to move data to reach AI.

“Sovereignty is our home court,” he said. “We do 70% of the world’s financial transactions. The most important data that travels around the world is in our Z systems. Those are sovereign systems. They’re on-prem, they’re in environments, they’re in banks, they’re in financial institutions, they’re in insurance. Sovereign is our home court and we know this space extremely well.”

The client-first philosophy extends to how IBM originally justified embedding AI into the Z processor — a bet made well before the ChatGPT moment. Financial clients told IBM they needed inline fraud detection at the point of transaction, not a report seven days later. That demand forced IBM to get aggressive about silicon-level AI before the broader industry caught up. The result is the model IBM intends to carry forward as inferencing specialization becomes the defining infrastructure challenge, according to Lewis.

“We had our financial clients coming to us saying, ‘Fraud is a billion-dollar industry, it’s a major problem. We need a way to be able to do that in line in the system,'” Lewis said. “They don’t invent for you — the clients — but if you listen really closely, they’ll tell you kind of what they really need and it forces you to get to that next place.”

Here’s the complete video interview, part of SiliconANGLE’s and theCUBE’s coverage of the Think 2026 event:

(* Disclosure: TheCUBE is a paid media partner for the Think 2026 event. Neither IBM, the sponsor of theCUBE’s event coverage, nor other sponsors have editorial control over content on theCUBE or SiliconANGLE.)

Photo: SiliconANGLE