1. Headline & intro
AI infrastructure has just been promoted from "critical for business" to "critical in war". Iran’s latest threats against the $500 billion "Stargate" AI data center project in the UAE make something uncomfortably clear: GPU farms are no longer just about training models, they are now strategic assets that can be attacked like oil refineries or undersea cables. In this piece, we’ll unpack what actually happened, why it matters for the global AI race, and why European companies, regulators and investors should pay close attention.
2. The news in brief
According to TechCrunch, Iran has warned it may strike major U.S.-linked energy and technology infrastructure across the Middle East if Washington follows through on threats to hit Iranian civilian facilities.
In a video released by Iran’s military and widely shared over the weekend, a spokesperson highlights potential targets and zooms in on the massive "Stargate" AI data center project in the United Arab Emirates, declaring that nothing is really hidden from Iran’s view, even if obscured on platforms like Google Maps.
TechCrunch notes that Stargate is a $500 billion joint venture between OpenAI, SoftBank and Oracle, announced in January 2025 to build huge AI data centers worldwide. The project reportedly struggled early on with funding and tariff-related costs but has been pursuing global expansion.
The latest warning comes amid a war that began in February and ongoing tensions around the Strait of Hormuz. TechCrunch reports that Iranian missiles have already hit Amazon Web Services (AWS) data centers in Bahrain and an Oracle facility in Dubai, and that Iran has also named Nvidia and Apple as potential future targets.
3. Why this matters
For years, "the cloud" has been sold as something abstract, geographically redundant and almost magically resilient. Iran’s threats — and the reported strikes on AWS and Oracle facilities — puncture that illusion. Racks of GPUs and power-hungry cooling systems sit in very real buildings, in very specific places, governed by very human politics.
The most obvious losers are the Stargate partners themselves. OpenAI, SoftBank and Oracle were betting that Gulf states could become the new backbone of global AI compute: abundant energy, friendly regulation, and proximity between Europe, Asia and Africa. Overnight, that bet looks riskier. Insurance costs will spike, financing for new sites will become more cautious, and customers will ask harder questions about where their "region" is actually located.
But the impact cascades far beyond one project. Any enterprise that chose Middle East regions for latency, cost or regulatory reasons now faces a new category of risk: physical disruption by state actors. For banks, logistics firms, or AI-heavy platforms running inference in these regions, continuity plans built only around power cuts and cyberattacks are suddenly outdated.
Competitively, this shifts momentum towards jurisdictions perceived as boring and safe: Northern Europe, parts of North America, maybe some Asian hubs. It also strengthens the argument for sovereign and regional clouds that keep sensitive workloads within politically aligned, militarily protected territories.
We are watching AI infrastructure cross a line: from commercial utility to strategic target.
4. The bigger picture
This is not the first time digital infrastructure has entered the crosshairs of geopolitics, but it may be the clearest example involving AI-specific assets.
Past incidents — from Stuxnet sabotaging Iranian nuclear centrifuges, to the Shamoon malware attack on Saudi Aramco, to ransomware that crippled Colonial Pipeline in the U.S. — showed that bits can hurt atoms. Undersea cable cuts and the suspected sabotage of the Nord Stream pipelines reminded Europe that invisible networks and energy lifelines are vulnerable.
What’s new here is the target class: hyperscale AI data centers. These are not just generic colocation halls; they are the physical bottlenecks of the AI economy. In a world where Nvidia GPUs are rationed and AI capability is increasingly seen as a component of national power, taking an AI data center offline is symbolically and practically closer to hitting a refinery than to defacing a website.
This sits atop broader trends: the weaponisation of supply chains (chip export controls to China), the concentration of AI capability in a handful of U.S. firms, and Gulf states trying to reinvent themselves as digital and AI hubs. The Stargate project was almost a caricature of this: U.S. AI IP, Japanese capital, U.S. enterprise software, Gulf location.
Compared with competitors, hyperscalers like AWS, Microsoft and Google have spent years quietly hardening their facilities. Yet even they have increasingly built in the Gulf to meet demand. Iran’s threats highlight that no amount of cyber-hardening compensates for geography. The industry is being forced to admit that "compute geopolitics" is now a real discipline, not a conference buzzword.
5. The European / regional angle
For Europe, this is a wake-up call on two fronts: dependency and proximity.
First, dependency. European companies may nominally choose an "EU region" for regulatory reasons, but many also use Middle East regions for redundancy or latency towards Asia and Africa. If those regions become contested, the promise of global failover weakens. Under the NIS2 Directive, critical entities in sectors like finance, health and transport must manage geopolitical and physical risks to their ICT infrastructure. Boards will have to justify why parts of their stack live within missile range of an active conflict.
Second, proximity. The Gulf is not far from EU borders in strategic terms. Disruptions to the Strait of Hormuz, as mentioned by TechCrunch, affect global shipping and therefore hardware supply for European data centers. Delays in GPUs, networking gear and transformers ripple directly into Europe’s own AI capacity.
At the same time, this creates an opening for European infrastructure players. Nordic data centers running on hydro and wind, continental hubs in Frankfurt, Amsterdam or Dublin, and emerging sites in Southern and Eastern Europe can pitch not just green power, but geopolitical stability. Projects like GAIA-X and Europe’s own exascale supercomputers suddenly look less like vanity projects and more like risk mitigation.
The EU has been talking about "digital sovereignty" for years. Iran just provided a brutal case study of what the lack of it can look like.
6. Looking ahead
Expect three shifts over the next 12–24 months if these threats persist or escalate.
First, risk models for cloud and AI infrastructure will be rewritten. Location will move from a procurement footnote to a board-level discussion. Multinationals will map their workloads against political fault lines the way they already map manufacturing sites against earthquake zones.
Second, AI data centers will increasingly be classified — formally or informally — as critical infrastructure. That means closer coordination with defense ministries, more physical hardening, and possibly even air defense coverage for the most strategic sites. The line between "civilian" and "dual-use" infrastructure will blur further.
Third, investors and regulators will re-evaluate mega-projects like Stargate. A $500 billion bet on concentrated compute in a geopolitically exposed location is harder to sell when the other side is openly publishing target imagery. Financing may shift towards a more distributed model: more, smaller sites across friendly territories, plus stronger edge computing to reduce reliance on a few mega-hubs.
For European readers, the key questions to watch are:
- Do EU regulators issue guidance on using cloud regions in active or high-risk conflict zones?
- Do hyperscalers quietly encourage EU customers to repatriate sensitive workloads to European regions?
- Does Europe accelerate its own GPU and data center build-out under the banner of resilience, not just innovation?
The opportunity is clear: turn Europe’s relative geopolitical stability into a selling point for AI infrastructure — before someone else’s missiles reroute your packets.
7. The bottom line
Iran’s threats against Stargate and other data centers show that AI infrastructure has entered the age of hard power. Compute is now a strategic asset, and its physical location matters as much as its API endpoint. For Europe, this is both a risk and an opening: a risk if we continue to rely on faraway, vulnerable hubs; an opening if we invest in resilient, sovereign AI capacity. The uncomfortable question every organisation should now ask is simple: who, exactly, is willing — and able — to defend your GPUs?
