Timothy Prickett Morgan

Sometimes, to solve a particular system architecture problem, you have to invent a new technology. And sometimes, you just need to squint at the problem a little and look at what you already have and use the parts in a different way.

There will probably never be a better time for any AI-related company to go public than between right now and next summer. The GenAI frenzy is at a fever pitch, and the big four hyperscalers and cloud builders alone – Amazon Web Services, Google Cloud, Meta Platforms, and Microsoft Azure – have collectively projected for capital expenses to be somewhere between $695 billion and $725 billion in 2026. There is probably at least that much expected to be spent between the big AI model builders (who are starting to build their own datacenters), hyperscalers in China, plus sovereign AI centers, HPC centers, academic centers, and governments who are also wanting to get in on the GenAI action.

The details are really thin, as usual, but IBM is still talking about the advantages that the Power10 and Power11 processors have over other server-class CPUs in the market today, as Power chief architect Bill Starke did at the most recent POWERUp conference in New Orleans. And Starke always ends his presentations with a hint or two about what Power Next or Power Future – what we would call Power12 – might look like.

We are all full aware of what the hyperscalers, cloud builders, and AI model builders are all doing to build out their datacenter infrastructure to support their AI aspirations. But it is hard to get a sense of what regular enterprises, service providers, governments, academic institutions, neoclouds, and sovereigns – who will eventually represent maybe half of total AI spending – are doing. So we have to use the sales of the traditional original equipment manufacturers as a kind of proxy.

The last time that Dell had a blowout quarter like the one that it has turned in was probably back when Michael Dell was selling custom PCs from his dorm room at the University of Texas. More than four decades later, the iron driving the business to new heights is, of course, AI systems, and is much bigger and more expensive but only delivering mid single digits of operating income. Which means Dell has to sell a lot more of it to get a return.

Nvidia wants the modern AI datacenter to be more like an Apple product, and with announcements it just made at the Computex conference in Taiwan, the company is getting closer to achieving that ambitious goal.

The GenAI waves are trying to lift all boats, and Hewlett Packard Enterprise is finally showing some profits as AI directly and indirectly boosts sales of AI systems based on Nvidia GPUs as well as traditional servers based on Intel and AMD servers. The company danced around its numbers, but as far as we can tell, sales of traditional servers far exceeded sales of GPU-accelerated machines in HPE's most recent quarter.

With DRAM and HBM memory makers running at full capacity and unable to boost capacity much, the one thing that will start slowing down is the constant upward revisions in AI spending and therefore global IT spending. There will be price increases that get passed through to customers, which will inch up the forecasts a little, but because HBM capacity is the gating factor in the AI accelerator market, there will not be leaps and bounds in revenue forecasts like we have seen every couple of months for the past several years.

In a world where applications and workflows are spread across multiple on premises systems and across the clouds, keeping track of how those applications and workflows are performing, fixing them when they stall or fail, and making sure everything is monitored and governing properly is a big job. And probably one that you don’t want to take on with so many other things going on.

There are a lot of things that are in short supply in the IT sector these days, and one of them is teams of engineers who can design complex chips and get the successfully through the fabs and packagers and onto system boards where they can be embedded into systems. Broadcom and Marvell, rivals in the chip shepherding businesses, are seeing their businesses explode thanks to this relatively new line of business, and it will not be long before this utterly dominates their business.