IBM’s Power Systems eyes AI inference boost amid AI-led enterprise growth

IBM’s Power Systems division remains waiting for the anticipated surge from GenAI, as overall infrastructure performs strongly and new Power11 hardware aims to bring AI closer to core business operations, potentially unlocking a new growth phase.

IBM’s Power Systems business is still waiting for the GenAI surge that many in the channel have been predicting, even as the wider infrastructure division delivered a strong start to 2026. In its first-quarter results, IBM reported Infrastructure revenue of $3.3 billion, up 15%, with Hybrid Infrastructure rising 28% to $2.09 billion and System z sales jumping 51%. The figures suggest the company’s high-end hardware cycle is doing some heavy lifting, but Power remains harder to isolate because IBM continues to bundle parts of the business with storage and other infrastructure lines.

Even so, the direction of travel appears positive. Industry analysis of IBM’s latest quarter points to overall revenue of $15.92 billion, up 9% year on year, while the software division remained the company’s main earnings engine. That matters for Power because much of the platform’s value now rests not just on hardware refreshes, but on the software and support stack that sits around it. IBM has also said it expects full-year constant-currency revenue growth of more than 5% and roughly $1 billion of extra free cash flow, underscoring confidence in the broader enterprise technology market.

The new Power11 generation is central to the case IBM is making. Marketing material for the platform says the systems are built to handle AI inferencing on the CPU itself, using Matrix Math Assist and related memory and acceleration features to run AI workloads alongside transactional applications without requiring separate GPU servers. That positioning is aimed at customers who want to bring inference closer to core business systems, rather than treat AI as an add-on in a separate environment.

Timing may also be helping the hardware cycle. A Power consultant quoted by TechChannel advised companies to review their systems annually and weigh support status, lease expiry and refresh timing, with the implication that newer hardware tends to be easier to defend operationally and commercially. That sort of guidance matters in a market where IBM has also been adjusting prices, and where memory and flash costs have climbed sharply enough to affect the economics of refresh decisions.

The bigger question is whether Power can benefit from the same kind of AI-led expansion IBM says it is seeing on its mainframes. In remarks on the company’s earnings call, chief executive Arvind Krishna described a future in which low-latency AI inference runs directly on core transaction systems, allowing banks and payment processors to apply models to every transaction rather than sampling a few. If that argument extends beyond System z to Power estates, IBM could yet find the GenAI bump it has been waiting for.

Source Reference Map

Inspired by headline at: ^[1]

Sources by paragraph:

Source: Fuse Wire Services