RAMmageddon Part 2: The geopolitical escalation — and why preparation matters even more now

Three weeks ago I wrote in my newsletter about the 2026 hardware crisis: exploding DRAM prices, server price hikes of 15 to 20 percent, CPU lead times of up to six months. The response was huge — and reality, unfortunately, has not contradicted me since. On the contrary: the situation has worsened. And the reason now lies not only in the fab, but increasingly in geopolitics.

The new dimension: when trade policy hits chip scarcity

When I wrote my last article, the main cause of the bottlenecks was clear: the AI industry is consuming 70 percent of global memory chip production, and less and less is left for the rest of us. That is still true — but now a second, massive disruption has been added on top.

On 14 January 2026, President Trump imposed a 25 percent Section 232 tariff on certain semiconductor imports into the US. For now it primarily affects a narrow category of AI chips (NVIDIA H200, AMD MI325X) that are being re-exported. But — and this is decisive — the proclamation explicitly leaves the door open for substantially broader tariffs on semiconductors, manufacturing equipment, and their derivatives. A report is due by April 2026.

In parallel: in December 2025, the US announced preparations for another round of tariffs on Chinese semiconductors — this time via Section 301. The tariffs are scheduled to take effect from June 2027, on top of the existing 50 percent on Chinese chip imports. And over all of it hangs Trump’s fundamental threat to slap tariffs of up to 100 percent on semiconductor imports if manufacturers don’t invest heavily in US production.

What does that mean for us in Europe? Three things simultaneously.

First: Global semiconductor supply chains are being further fragmented by tariffs and export controls. Manufacturers have to reallocate capacity, adjust shipping routes, comply with new compliance requirements. That costs time and money — and both get passed on to us.

Second: Europe is at the back of the queue. The US is securing preferred access to chip capacity through framework agreements with Taiwan, South Korea and Japan. Europe has not signed deals of comparable depth. The EU Chips Act mobilizes 43 billion euros, but new fabs take years before they’re delivery-capable. For the foreseeable future Europe remains structurally dependent on advanced semiconductors — TSMC alone holds 65 to 70 percent market share on chips under 7 nm.

Third: The Taiwan risk. An analysis from the Sovereign Europe Forum at the Munich Security Conference 2026 puts it bluntly: a blockade or military escalation around Taiwan would hit the global economy massively. For Europe it would not be a side event but a direct prosperity and stability shock. The risk itself is not new — but combined with the current chip scarcity and the tariff escalation, it becomes a factor IT leaders must include in their planning.

The war in Iran: the supply chain crisis gains a new front

And as if all of that weren’t enough, since 28 February 2026 another dramatic dimension has opened up: the armed conflict in Iran.

What looks at first glance like an energy story hits our IT sector directly. The Strait of Hormuz — through which roughly 20 percent of the world’s oil is transported — is effectively blocked. Iran has stopped transit, Maersk has suspended all shipping through it, freight aircraft from the region are grounded.

But it isn’t only about oil. The same sea routes carry semiconductors and batteries from Asia, pharmaceuticals from India, and petrochemical precursors for electronics manufacturing. Cargo ships are stuck in the Gulf or taking the massive detour around the Cape of Good Hope — with corresponding delays and skyrocketing shipping costs.

Patrick Penfield, professor of supply chain management at Syracuse University, sums it up: the longer the conflict drags on, the more likely shortages and massive price increases become across the whole breadth of supply chains.

For our industry that means: the already strained hardware supply is being further compressed by logistics disruptions. Anyone hoping for fast delivery of new servers, memory components or networking hardware has to expect even longer lead times. At the same time, rising energy prices push up data center operating costs — another reason to maximize the efficiency of existing infrastructure instead of waiting for new hardware.

The picture in March 2026: a perfect storm

Let’s sum up what has stacked up since my last newsletter.

Section 232 tariffs on semiconductors have been in force since 15 January 2026 — and can be extended to broader product categories at any time. Intel and AMD are reporting supply shortages on server CPUs, particularly in China — with lead times of up to six months and price uplifts of 10 percent or more. Memory manufacturers have no plans to increase production capacity for standard DRAM. Samsung, SK Hynix and Micron continue to prioritize HBM and AI chips. Export controls on critical minerals are on the horizon — in October 2025 China announced export restrictions that are currently suspended until November 2026. And now: the Iran conflict is blocking one of the world’s most important trade routes and hitting semiconductor supply chains from Asia, while rising energy prices simultaneously push up data center operating costs.

The structural redistribution I described in my last article is not just being cemented by the geopolitical escalation — it is being amplified at an entirely new level by the war in the Middle East. What began as chip scarcity has now become a perfect storm of supply restrictions, tariffs, export controls and logistics disruptions.

What’s changing in my recommendations — and what isn’t

My base formula from the last newsletter still stands.

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

But the geopolitical dimension adds two important elements.

Data sovereignty becomes a must, not a nice-to-have. When global hardware supply chains fragment, we have to think even harder about where our data and workloads live. “Made in Germany / Europe” is not just a quality seal — it is strategic insurance. Anyone running their critical workloads with a European provider is less exposed to decisions made in Washington or Beijing.

We at CID live exactly that: German data centers, European value creation, full data sovereignty.

Vendor diversification becomes strategic. In a world where tariffs, export controls and capacity bottlenecks can restrict the availability of specific components at any time, betting on a single hardware vendor or single chip architecture becomes risky. In your POCs, deliberately test alternative platforms, open standards, and software-defined approaches that give you flexibility in hardware choice.

The silver lining: finally thinking sustainably — because economics is forcing us to

I know this so far sounds like a bleak outlook. But I want to take a perspective that may surprise: this crisis is forcing us to do something we should have been doing long ago.

Let’s be honest: the usual rhythm — replace hardware every three to five years because support runs out and the depreciation is through — was comfortable but rarely sensible. The hardware that regularly flew out was 90 percent still fully functional. The technology wasn’t the problem, the support risk and the habit were.

Now that new hardware is unaffordable or simply unavailable, something interesting is happening: we are being forced economically to think sustainably. And that — honestly — was overdue.

The usual time-sink projects of “hardware out because depreciated and no support” disappear right now. And that is exactly where the real opportunity lies: we suddenly have both the time and the necessity to address causes instead of just treating symptoms.

The real lever: modernize software instead of replacing hardware

When we modernize the software that runs on the hardware and rebuild it cloud-native, the requirements and architectures for hardware change dramatically. Suddenly you don’t need the biggest machine with the most RAM — you need the right architecture for the actual workload.

This is the moment IT organizations should ask themselves honest questions. Questions there was never time for in the normal refresh cycle.

If software or databases are inherently redundant and resilient — why do the hardware and the operating model beneath also have to be? How much expensive hardware redundancy are we operating only because the software running on top was never built to be fault-tolerant itself? If we modernize the software, we can radically simplify the hardware requirements.

Why must “AI” always run on GPUs? The reflex “AI = GPU” is an expensive misconception in many cases. The truth is: a lot of what gets sold under the “AI” label is solvable with classic process digitization and automation — perhaps complemented with one small, targeted AI component. And that component often runs perfectly well on standard CPUs, not on a 40,000-euro GPU that isn’t deliverable anyway.

I’m not saying GPUs don’t make sense — for training large models and certain inference workloads they are essential. But for 80 percent of what companies today define as an “AI project”, there are smarter, cheaper and — especially in the current situation — more realistic paths.

All these questions should be discussed and addressed NOW. Not in a year when new servers arrive and the old groove continues. Now is the time to apply some brain work, to question the existing, and to allow ourselves new thinking.

CID: we are using the time — with and for our customers

We at CID were lucky to have finalized our hardware orders together with our CEO Rouven Lörch back in September 2025 and to have placed them in 2025. Our capacity is in place, and we’re now using it intensively — for POCs with our customers, for architecture tests under real load, for validating new platforms.

What we see in those tests confirms exactly this: anyone who applies right-sizing consistently, modernizes their software, and uses AI in a targeted — not blanket — way can extract significantly more from existing hardware than expected. That is not a consolation prize — it is a real lever that works independently of lead times and tariff escalations.

Europe is running into a challenge — but also into an opportunity

The combination of AI-driven demand, geopolitical fragmentation and tariff escalation makes 2026 the most demanding year for hardware procurement since the Covid crisis. And unlike then, there is no clear end point.

But that is precisely why we have to stop solving the problem only on the hardware side. The real transformation happens in software, in processes, in architectural thinking. And for that transformation, the crisis has paradoxically given us what we never had: time.

The extended formula is:

Extend support to gain time. Modernize software and think cloud-native — that changes hardware requirements fundamentally. Question existing architectures: do we need hardware redundancy, or is software resilience enough? Evaluate AI honestly: what truly needs a GPU — and what is smart process digitization? Run POCs and MVPs to test the future. Order with precision when the market allows — not as much as possible, but as much as necessary. Set data sovereignty and vendor diversification as strategic guardrails.

Whoever plans today buys better tomorrow. Whoever orders in panic today pays twice. And whoever modernizes their software today may need significantly less hardware tomorrow than they thought.

---

Sources: White & Case (“President Trump orders 25% Section 232 tariff on advanced semiconductors”, Jan 2026); Tax Foundation (“Tariff Tracker: 2026 Trump Tariffs & Trade War”); Tom’s Hardware (“Trump administration announces new tariffs on Chinese chips”); Tom’s Hardware (“Intel, AMD server CPUs suffering from supply shortages”); Springer Professional (“Diese fünf Engpässe bestimmen die Chipverfügbarkeit”); United Europe / MSC 2026 (“Sovereign Europe Forum”); Schaaf Media (“Ausblick 2026 — Geopolitische und Wirtschaftliche Dynamiken”); All-Electronics (“Nexperia und Europas Abhängigkeit von globalen Lieferketten”); AP News / Mercury News (“Supply chain disruptions from the Iran war could raise prices for drugs, electronics and more”, March 2026); PBS NewsHour (“Experts analyze what the Iran war could mean for U.S. gasoline prices”, March 2026).