A Tale Of Two Server Markets

January 12, 2026 Timothy Prickett Morgan

Happy New Year! It is hard to imagine what might happen next. But what I can tell you for sure is that until the gas comes out of the GenAI expansion, infrastructure spending in the datacenter in the aggregate will continue to be increasingly dominated by AI systems even if this is not the case at most IBM i shops.

Before the Dot Com Boom, the market researchers at Gartner and IDC put out competing reports on sales of servers and external storage arrays, which we found very useful and you no doubt did, too. In 2022, both companies stopped giving the high level of their complex server and storage datasets out to the public, much to our chagrin.

But IDC has been putting out a report every once in a while in recent years, which gives us a chance to fill in a few gaps and also make some comparisons to server spending during the Dot Com Boom, the bust that followed it, the climb after that, the rapid decline during the Great Recession, the collapse of the RISC/Unix markets and the steady decline in revenues for various kind of proprietary systems over the past three decades. I did a complicated analysis of the historical IDC data, the company’s sporadic updates, and its jaw-dropping forecasts out to the end of 2029 in The Next Platform in mid-December when this data came out. I am not going to redo all of that work here, but I am going to ask you to take a look at the charts in that story just so we can put all of this nonsense talk about how this is like the Dot Com Boom.

What we are experiencing is totally freaking crazy. It took the server market decades to get into the $12 billion to $13 billion per quarter range in a sustained fashion, and it took the first wave of machine learning-based AI spending to get it above the level of $15 billion per quarter in server spending during the peak of the Dot Com Boom. During the coronavirus pandemic, GenAI was starting and clouds and hyperscalers were spending huge amounts on infrastructure because everybody in the world started working remotely, and that drove the server market to a very wiggly average of around $20 billion per quarter.

In 2025, thanks to the GenAI boom, the average sales per quarter will be somewhere north of $100 billion, and the forecast from IDC out through 2029 implies an average that will be on the order of $150 billion per quarter; based on what AI model makers are doing, it could be $200 billion per quarter.

Even when you adjust those numbers for inflation, the amount of quarterly IT spending on servers is easily a factor of 5X to 6X over those three decades, and most of the growth will be happening between 2025 and 2029 if you believe the forecasts from IDC.

We live in a Power Systems business that is growing at a few points per year, and in a good year, like 2025, is struggling to push somewhere on the order of $1.67 billion in iron to customers that are absolutely dedicated to their back office systems and the platforms that run them. As I reported back in October, I had expected $2 billion in revenues for Power Systems in 2025, but unless the fourth quarter turns out different than we expect, that is not going to happen. Still, there are operating systems, middleware, databases, development tools, databases, and tech support that are sold on top of this, so don’t think this is a small or unprofitable business for Big Blue.

Power is very, very important to IBM, and will continue to be, for this very reason. And IBM is determined to sell customers its own AI software stack (which includes licensed models from Anthropic) and its own Spyre AI accelerators to help run inference for those models as they are integrated into the applications and databases running on Power iron. I stand by the statement that by doing this, IBM can double the Power Systems business.

This is the year we might start seeing this happening. We think that even skeptical IBM i shops know they need to do something with GenAI, and they will widen their budgets to do this, just the hyperscalers, cloud builders, AI model builders, and large enterprises have done to cause that exponential increase in server spending outlined by IDC and massaged by me over at The Next Platform.

What is your company going to do? What’s your budget, and what’s your plan? Let me know.

This Issue Sponsored By

Table of Contents

Content archive

Recent Posts

Subscribe

Pages

Search