The global boom in artificial intelligence is often portrayed as a race between software companies and semiconductor designers. But the reality is far more physical. Artificial intelligence is not merely code—it is infrastructure. It requires massive data centers, constant electricity, rare industrial gases, semiconductor fabrication plants, and global fiber-optic cable networks capable of moving extraordinary amounts of data across continents.
In effect, the AI revolution is the largest infrastructure expansion program in human history. Technology companies are projected to spend hundreds of billions of dollars annually building hyperscale data centers and purchasing advanced GPUs to power the next generation of AI models. But this enormous buildout rests on fragile physical systems that depend on energy supply chains, industrial materials, and global telecommunications networks.
The war involving Iran has exposed just how vulnerable these systems are. What began as a regional military conflict is now rippling through the infrastructure that supports the global digital economy. Energy markets have been disrupted, critical semiconductor inputs have been constrained, major internet cable corridors have been compromised, and for the first time in history hyperscale data centers themselves have become military targets.
Taken together, these disruptions threaten to slow—or fundamentally reshape—the global AI data center boom.
1. The Energy Shock: AI’s Hidden Dependence on Natural Gas
Artificial intelligence has a power problem.
Training and operating large AI models requires enormous quantities of electricity. Hyperscale AI campuses can consume multiple gigawatts of power—comparable to the energy demand of a medium-sized city. The next generation of AI clusters, designed to support hundreds of thousands of GPUs, will push power demand even higher.
While renewable energy capacity is expanding rapidly, most AI infrastructure today still depends on reliable baseload electricity. Data centers cannot shut down when the sun sets or when wind output fluctuates. Training runs lasting weeks or months require uninterrupted power.
For this reason, natural gas has become the default energy source for many new AI facilities. Gas-fired power plants can operate continuously, ramp production quickly, and can be built faster than nuclear reactors or large-scale renewable installations. As a result, a growing number of data center projects are being paired with dedicated gas power plants.
This creates a direct link between the AI economy and global natural gas markets.
The war in Iran has placed that energy system under stress. The Persian Gulf is a critical corridor not only for oil but also for liquefied natural gas shipments. Instability in the region raises insurance costs, disrupts shipping routes, and creates uncertainty around future supply.
When natural gas prices rise, electricity prices follow. For AI companies, energy costs are one of the largest operational expenses. Spiking electricity prices can reduce the profitability of AI training runs, increase inference costs for deployed models, and delay the construction of new data center campuses.
The AI industry, in other words, is discovering that its growth is tied not just to silicon and software but also to pipelines, LNG terminals, and global energy geopolitics.
2. The Helium Shortage: A Hidden Threat to Semiconductor Manufacturing
Energy is not the only vulnerability in the AI supply chain. Another critical dependency lies in a gas that rarely receives public attention: helium.
Helium is essential to semiconductor fabrication. Ultra-high-purity helium—known as 6N-grade helium—is used to cool silicon wafers during etching in advanced chip manufacturing equipment. Without it, modern semiconductor fabrication simply cannot function.
Qatar produces roughly 30 percent of the world’s helium supply, much of it from the massive Ras Laffan industrial complex. This facility is one of the few locations on Earth capable of producing helium at the purity levels required by advanced semiconductor fabs.
The war has disrupted this system. Drone strikes on Ras Laffan infrastructure forced production shutdowns and triggered a global helium shortage. Semiconductor manufacturers in South Korea and elsewhere have already implemented conservation protocols to stretch their remaining supply.
The consequences are significant. Every advanced GPU powering an AI training cluster—from NVIDIA accelerators to custom AI chips—requires semiconductor fabrication that depends on helium. If helium shortages persist, chip production will slow. And if chip production slows, the pace of AI data center expansion will slow as well.
This link illustrates the fragility of the AI supply chain. A conflict thousands of miles from Silicon Valley can constrain a rare industrial gas, which in turn affects the production of chips that power artificial intelligence systems worldwide.
3. The Internet Chokepoint Crisis: 90% of East–West Data Traffic at Risk
The most serious threat to the AI infrastructure boom, however, may lie in global telecommunications networks.
Most people understand the Strait of Hormuz as an oil chokepoint. What is less widely understood is that the region also sits at the center of the world’s digital connectivity.
Multiple submarine cable systems pass through the Persian Gulf and the Strait of Hormuz. Even more passed through the Red Sea. These cables form the primary digital bridge between Europe and Asia, carrying most of the intercontinental internet traffic.
Together, these corridors carry roughly 90 percent of east–west data capacity.
For the first time in modern telecommunications history, both routes have been effectively compromised simultaneously.
Military activity, shipping disruptions, and insurance restrictions have turned the Persian Gulf and Red Sea into high-risk zones for cable operations and repair ships. When cables fail or require maintenance, specialized vessels normally perform repairs within days or weeks. In a war zone, those repairs may not happen for months.
As a result, internet traffic between Europe and Asia is being rerouted through longer paths across the Pacific Ocean and North America.
This dramatically increases latency. Direct routes that normally take around 130 milliseconds now exceed 250 milliseconds as traffic circles the globe.
For ordinary internet users, this may mean slower connections or occasional disruptions. For artificial intelligence systems, however, the consequences are far more severe.
AI infrastructure increasingly depends on distributed computing. Training runs may involve thousands of GPUs spread across multiple data centers in different regions. These clusters must exchange massive amounts of data with minimal latency.
When latency doubles and bandwidth capacity shrinks, distributed AI training becomes significantly less efficient. In extreme cases, it may become impractical.
Bandwidth constraints pose an even greater problem. The surviving trans-Pacific cable routes simply do not have enough spare capacity to absorb the loss of major Europe–Asia corridors without congestion.
This means the AI boom now faces a bottleneck not in computing power but in global network infrastructure.
4. Data Centers Become Targets: A New Era of Infrastructure Warfare
Perhaps the most dramatic shift triggered by the Iran conflict is the emergence of data centers as military targets.
In early 2026, Iranian drones struck several hyperscale data center facilities in the Gulf region, including infrastructure associated with major cloud providers. The attacks damaged power systems and cooling infrastructure and forced operators to shut down parts of their operations.
This marked the first confirmed kinetic military attack on a hyperscale cloud data center.
The implications are profound.
For decades, data centers were treated as purely commercial infrastructure. Even during major conflicts, telecommunications networks were rarely targeted directly. But the role of cloud computing in modern economies—and increasingly in military systems—has changed that calculation.
Cloud infrastructure now supports everything from banking systems to logistics networks to military data analysis. As a result, it has become strategically valuable.
The attacks demonstrated that hyperscale data centers are physically vulnerable. Cooling systems, electrical substations, and backup generators are large, exposed pieces of infrastructure that can be disabled with relatively inexpensive weapons such as drones.
This creates a new risk calculus for the AI industry. Building a $20–$30 billion data center campus in a geopolitically unstable region suddenly looks far less attractive if that facility could become a wartime target.
Insurance costs, security requirements, and geopolitical risk assessments will now influence where future AI infrastructure is built.
The AI Boom Meets Geopolitics
The war involving Iran has revealed a fundamental truth about the artificial intelligence economy: it is built on physical systems that exist in the real world.
AI depends on energy systems that can be disrupted by geopolitical conflict. It depends on rare industrial materials such as helium that can become scarce overnight. It depends on submarine cable networks that pass through narrow maritime chokepoints. And it depends on massive data center campuses that can be targeted during war.
For years, the technology industry assumed that its supply chains were insulated from geopolitics. The digital economy seemed weightless, detached from the physical world.
The war in Iran has shattered that illusion.
Artificial intelligence may be the most advanced technology ever created. But the infrastructure that supports it remains deeply vulnerable to the oldest forces in human history: geography, conflict, and scarcity.
The AI boom is not ending. Demand for computing power will continue to grow, and companies will keep building data centers around the world.
But the conflict has changed the landscape.
Where those data centers are built, how they are powered, how they connect to global networks, and how secure they are from geopolitical disruption will now matter as much as the algorithms running inside them.
The economic consequences of derailing AI data center buildouts and operations are massive – far more significant than the energy shock caused by this war. This has not only derailed progress but will also cause significant delays as China continues unabated with its own AI infrastructure build-out. The long-term impact of this will also be profound.
The next phase of the AI race may therefore be decided not only in semiconductor fabs and research labs—but also in shipping lanes, energy markets, and contested choke points like the Strait of Hormuz.
Leave a Reply