Perhaps 2026 Is The Year For Power Systems To Boom A Little

February 2, 2026 Timothy Prickett Morgan

In wrapping up IBM’s call with Wall Street analysts in the wake of Big Blue reporting its financial results for the fourth quarter of 2025 – which was a very good one for the most part, by the way – chief executive officer Arvind Krishna referred to the IBM Company as a “software-led, platform-centric company.”

This is no news to any of us who have been using or programming or analyzing the business of IBM over the many decades. Of course this is true. But for IBM, the main platform that is driving the big bucks at the top and bottom lines is the System z mainframe. And the reason is simple: The System z line probably drives a factor of 2X more revenues over its also popular Power Systems line.

But once again, neither Krishna nor chief financial officer Jim Kavanaugh said “Power” or “Power Systems” and gave very little insight into how the Power Systems business is doing as 2025 came to an end other than to talk very generally about the Distributed Infrastructure division that combines Power Systems with all of IBM’s tape, disk, and flash storage products. IBM might was well call this “Other Stuff,” and be perfectly vague about it. For one thing, for most Power Systems shops, the machine is their main compute – literally, their main frame – and for all intents and purposes a mainframe to run databases and transaction processing for systems of record that encapsulate the functions of the businesses that use platforms based on Big Blue’s Power processors. These mini-mainframes are different from System z machines in so many ways, but functionally equivalent.

The main fault with the Power Systems business is that its software content (outside of the IBM i platform and meaning mostly AIX) is lower than on mainframes, which have very expensive hardware and even more expensive monthly rental fees for software that racks up pretty high. This fact is why nearly half of IBM’s revenues now come from software – including operating systems, transaction processing, middleware, database, AI frameworks and models, applications, and development tools – and drives two-thirds of its profits.

The hardware is a necessity if IBM wants to sell software, but it is inherently difficult and expensive to develop. There will come a day, hopefully more than a decade from now, when Big Blue will be forced by the economics of chip development to further converge its System z and Power platforms, or sell it off (likely to Hitachi) and have it add capacity at an even slower pace than it now does.

But that day is not today. And even should that day come, IBM (or a third party) will support that hardware and its software for a long time after that day comes. None of us that have a little gray in our beards need worry – we will retire before IBM does.

In the meantime, let’s count the IBM systems money, which is still sufficient to justify the investment in two distinct platforms that nonetheless share a lot of functions.

In the fourth quarter, IBM brought in $19.69 billion in revenues, up 12.2 percent and had a pre-tax income of $11.93 billion, up 14.3 percent, and a net income of $5.6 billion, nearly double from a year ago. IBM exited the quarter with $14.47 billion in cash, which gives it plenty of maneuvering room if you ignore the massive debt it has managing the portfolio of gear it has in the field with resellers and under lease with customers and the money it is still paying down for the $34 billion acquisition of Red Hat back in July 2019. If you add up the collective revenues from Red Hat since then, the software company has brought in over $40 billion in revenues. So, in that sense, IBM has gotten its bait back.

IBM’s Infrastructure group, which sells servers, operating systems, storage, and occasionally switching as well as tech support and secondhand equipment, had $5.13 billion in sales, up 20.6 percent year on year and up 44.2 percent sequentially from a pretty weak Q3 2025, which is when Power11 machines launched and when System z17 machines were still ramping.

Servers and storage – what IBM mysteriously calls “Hybrid Infrastructure” – accounted for $3.84 billion in our model, up 29 percent, while Infrastructure Support was up 1 percent to $1.29 billion.

What we don’t know is how the Power Systems line did, but we do know that Distributed Infrastructure (Power Systems plus storage) had sales up 3 percent as reported and flat at constant currency. When we plug these numbers in our model and do some spreadsheet hocus pocus guessing how IBM did in storage in general and with Power-based storage arrays in particular, then we reckon that Power Systems revenues were up 6.9 percent to $600 million.

This was pretty good growth, but not enough for Power Systems to break through $2 billion in annual sales, according to our model, and we blame the extra time it has taken IBM to put Power11 systems in the field as well as the “Spyre” AI accelerators that could turn out to be the hardware for the code assistant killer app that might evolve on the IBM i and AIX platforms.

By our math, for the full year, we think Power Systems overall – including capacity rentals on the Power VS cloud – came to $1.61 billion, up 1 percent. We admit that we are less sure about the revenue coming from the hundreds of Power Systems customers who are renting capacity on the IBM Cloud through the Power VS offering. It is not really material enough for Krishna or Kavanaugh to talk about. Heck, they don’t even talk about the much bigger IBM Cloud, which has over five dozen datacenters around the world doing all kinds of things for customers. IBM used to give some

As we have pointed out many times before, IBM is one of the world’s largest platform providers in the world, has been for six decades, and will remain one for the foreseeable future. If you strip it down to its basics – servers, storage, switching, operating systems, tech support, and financing – we think this underlying business is very strong and supports that software business that Krishna and Kavanaugh talk about all the time. Without it, there is no software business. And without that sticky software, which is entangled into the back office functions of perhaps over 100,000 customers worldwide who still spent money with IBM on systems (and perhaps half again as many if you include those who run on vintage hardware and software), there would be no need for ever-embiggening hardware. They go together, like yin and yang, chocolate and peanut butter, Tom & Jerry.

In Q4 2025, we think this “real” systems business at IBM brought in $9.42 billion in sales, up 17.3 percent, and had a pre-tax income of $5.18 billion, up 21 percent and comprising 55 percent of revenues. This is the largest revenues for “real” systems since Q4 2019 in our model, and the current business is considerably more profitable now than it was back then. But, such spikes happen, as they did in Q4 2014 and to a lesser extent in Q4 2017. This is not the norm. But, we will take it.

Perhaps, as System z “wraps around” against the exceptional growth in 2025 to more normal sales levels in 2026, it will be the turn of Power Systems to have an exceptional year. We remain hopeful, but we shall see. It is too early to make any predictions.

This Issue Sponsored By

Table of Contents

Content archive

Recent Posts

Subscribe

Pages

Search