Storage crisis and hyperscalers: Until the competitors in Europe have starved

First high, then outrageous hardware prices: Together with the AI hype, the cost spiral for data center equipment is turning faster and faster. The storage crisis is good for manufacturers but bad for most customers. For iX author Daniel Menzel, however, it doesn't stop at numbers – he learns every day how dramatic the situation in data centers really is.

Daniel Menzel english

Daniel Menzel is the managing director of Menzel IT GmbH in Berlin and works with his team to build HPC, ML, and private cloud computing clusters.

The numbers on prices and missed deliveries are one thing. But how dramatic is the situation really in the local data centers?

For existing projects, it has hardly been problematic so far. Of course, you grit your teeth when you suddenly have to pay 5 times the normal price for RAM upgrades. But for expansion and new projects, the situation is, in my opinion, nothing less than disastrous. I'm aware of tenders that are currently being completely withdrawn. Entire million-euro projects that are indefinitely put on hold. And also actual upcoming legal disputes because suppliers have given a price guarantee and customers are now demanding it. This initially affects “only” the hardware, but of course has a knock-on effect because the hardware is needed for things. The situation is indeed currently... grotesque.

Which areas are particularly impacted?

Uff... I would say, actually all of them. Trivial answer: Of course, the areas where more hardware is used are also more affected by price increases. I'm thinking particularly of HPC and AI, but also (European) cloud providers who want to grow. This seems particularly painful for the latter, as they don't have a large margin in their competitive relationship and therefore have to calculate prices very tightly.

If it doesn't have to be the latest and most performant hardware: What about the used market?

Personally, I'm not a big fan of used hardware in racks. That certainly comes from my high-performance mindset: we always run our systems at the performance limit. And that extends to cloud/enterprise systems: if I procure really high-performance nodes there, they might cost twice as much initially, but they deliver three times as much workload (and thus money). Used systems only bring me a higher electricity bill.

Nobody likes to deal with a bad situation: What workarounds or detours have helped so far?

Currently, I have the feeling that there is a collective agreement in Europe to bury our heads in the sand – occasionally paired with loud and sometimes fatalistic lamentations that Europe has made itself too dependent on others in the past 20 years and will never get out of it again. There are no real approaches yet. However, one must also consider: Prices have risen sharply in the past four months. That is a relatively short time for investments. You can certainly postpone a new project a little if the outflow of funds is not absolutely necessary in 2025 (funding, public sector, ...).

Hyperscalers in particular can afford the higher prices. Are we being forced into even greater dependence?

Yes. No, wait: Yes!!! Hyperscalers have large financial reserves and are used to speculating on the future. They calculate for their investors how long it will take for the small competitors in Europe to starve. In other words, when the bottom has been reached and when the break-even point is subsequently achieved. And above all: how large the ROI will be in five years. If these numbers are sufficiently good, the cash flow for hyperscalers feels almost infinite.

I certainly wouldn't go so far as to say that hyperscalers have engineered this situation to their advantage, but I suspect that people on the other side of the pond are not entirely unhappy about it either.

Mr. Menzel, thank you for your answers!

(fo)