RAMmageddon Part 1: Hardware renewal 2026 — time for preparation, not panic orders

In my last post I talked about our tech refresh and our plans for 2026. But since the start of the year the hardware market has deteriorated dramatically — and that will hit every infrastructure project this year directly. Anyone who has to order server hardware right now is facing a bitter reality.

The facts: a crisis that has earned its name

What we’re experiencing right now has been dubbed “RAMmageddon” — and the name fits.

Samsung has raised its prices for server DRAM (DDR5) by up to 60 percent since September 2025. DRAM contract prices have risen by roughly 170 percent over the course of 2025. Dell, Lenovo, HP and HPE have announced server price increases of 15 to 20 percent — some already effective since December 2025. Intel is warning of CPU lead times of up to six months. SK Hynix reports: HBM, DRAM and NAND capacities are sold out through the end of 2026. IDC, in its pessimistic scenario, forecasts a market drop of up to 9 percent.

Dell COO Jeff Clarke puts it bluntly: “This is the worst shortage I have ever witnessed.”

The reason? AI infrastructure is eating everything. Up to 70 percent of the world’s memory chip production in 2026 will flow into hyperscaler data centers. Manufacturers like Samsung, SK Hynix and Micron have massively reallocated production capacity towards high-margin HBM and AI chips — at the cost of “normal” DDR5 and DDR4 for enterprise servers. Micron even retired its 30-year-old consumer brand Crucial to focus entirely on the AI market. Silicon Motion CEO Wallace C. Kou sums it up: “We’re facing what has never happened before: HDD, DRAM, HBM, NAND… all in severe shortage.”

This is not a temporary dip — it is a structural redistribution of global semiconductor capacity.

A Covid-effect on the data center

The parallels to the Covid-era chip crisis are not exaggerated — the Wall Street Journal is already drawing them itself. But this time the cause is not a pandemic, it is an AI-driven demand shock that hits all segments simultaneously: DRAM, NAND, HDDs, CPUs, even glass fabric for chip production.

For IT organizations that planned hardware refreshes in 2026, that means concretely: longer lead times, shorter quote validities, ballooning budgets — and very difficult conversations with the CFO.

Support extensions: a sensible bridge, not a stopgap

My first recommendation is pragmatic: support extensions on existing systems are not just a stopgap right now — they are a strategically sound move. They buy you time without forcing you into overpriced orders.

But — and this is the decisive point — that gained time has to be put to active use.

The actual opportunity: planning for things we never had time to plan for

When was the last time IT had the calm to rethink existing architectures and software solutions from the ground up? Essentially never. It was always: old system out, new system in, lift-and-shift, keep moving.

That is exactly where the opportunity in the crisis lies.

Proofs of concept and minimum viable products — now.

Use the forced pause to work intensively on the future architecture. Test whether technologies like virtualization, KubeVirt, modern SDN concepts or containerized platforms suit your workloads. Analyse where over-dimensioned infrastructures are generating cost — not only at VMware, but across the entire software landscape.

We at CID are doing exactly that: we’re currently running POCs across several fire compartments, testing under real load, and gathering insights no datasheet can supply.

And please: no panic lift-and-shift into the cloud

I can hear it already: “Then we’ll just go to the cloud.” But an unprepared lift-and-shift is not a solution — it is a cost trap. Anyone who lifts their over-dimensioned on-prem infrastructure 1:1 into the cloud ends up paying more than for overpriced servers. The inefficiencies just travel along — only now you pay for them monthly instead of once.

Cloud migration makes sense — but only with a thought-through architecture, right-sizing, and a clear strategy. And that is exactly what you can develop now.

Preparation is the best alternative

When prices calm down again — and at some point they will — you will be ready. You order what is actually needed. Not as much as possible, but as much as necessary. That is the actual right-sizing approach.

The formula is:

Extend support → rethink architecture → run POCs → order with precision.

Anyone who orders in panic now pays premium prices for an architecture of yesterday. Anyone who plans now invests in an architecture of tomorrow — at the prices of tomorrow.

---

Sources: Tom’s Hardware (“Data centers will consume 70% of memory chips in 2026”); TrendForce (“Dell Hikes Prices 15-20%, Lenovo from January 2026”); Tom’s Hardware (“Intel, AMD server CPUs suffering from supply shortages”); Network World (“Server memory prices could double by 2026”); TechWire Asia (“Memory chip shortage 2026 reaches crisis levels”); Dataconomy (“Global Memory Chip Shortage to Drive Up Tech Prices in 2026”); Wikipedia (“2024–present global memory supply shortage”).