How The Iran Conflict Is Threatening AI’s Power Grid

Welcome to Memorandum Deep Dives. In this series, we go beyond the headlines to examine the decisions shaping our digital future. 🗞️

This week, we’re examining a vulnerability the AI industry rarely discusses: the physical and geopolitical infrastructure that keeps its systems running. For years, the conversation around AI risk focused on algorithms, regulation, and safety. But the rapid escalation of conflict in the Middle East has exposed a different kind of fragility, one rooted in energy grids, data centers, submarine cables, and the global flows of capital that finance the AI boom.

When drone strikes damaged Amazon Web Services facilities in the Gulf, the incident served as an early signal that AI’s infrastructure is no longer insulated from geopolitical conflict. What once looked like a purely technical race—build bigger models, secure more chips, scale more compute—is increasingly shaped by older forces: war, energy markets, and strategic geography. As the AI economy expands, its most important bottlenecks may no longer be inside the model, but in the world around it.

The Hidden Backbone of AI

Ever since the end of the Cold War in the early 1990s, global economic structures have evolved as an interconnected web that relies on the continued movement of goods and services to operate smoothly. While this has been enormously profitable for enterprises and consumers, this interconnectivity also means that problems in one part of the world can trigger a domino effect, with repercussions felt around the globe.

For the AI industry, this meant ensuring a steady flow of chips, power, and human capital to sustain its operations as it continued to scale. Over the past few years, the industry has worked on planning for chip shortages, model safety, and regulatory pushback. What it failed to plan for was a conflict that could threaten energy costs and data flow, both of which are essential components of running AI systems.

In March 2026, the lack of planning led to real-world consequences when strikes hit three Amazon Web Services data centers across the United Arab Emirates and Bahrain. The attacks caused structural damage, knocked out power, triggered fire suppression systems, and flooded server equipment. The result was disruptions in the functioning of banking apps, ride-hailing, and enterprise software, and a call from AWS for its customers to migrate their workloads out of the Middle East.

For the broader tech industry, it was a signal of the days ahead, when physical infrastructure would prove the weak link. The U.S.-Israel strikes on Iran, under Operation Epic Fury, triggered a chain of consequences that reaches far beyond the battlefield.

Data centers under fire

Within days of the U.S. and Israel’s coordinated strikes on Iran, retaliatory drone attacks targeted AWS data centers. What made these retaliatory strikes especially significant was the rationale provided by Iran’s state media, which claimed that Bahrain’s AWS facility was targeted because it hosts U.S. military workloads.

The claim was not without merit, as the U.S. military reportedly ran Anthropic’s Claude AI on AWS infrastructure to support intelligence assessments and targeting during the Iran strikes. These capabilities run on the same commercial cloud that powers banking and customer services, making them prime targets for adversaries.

Data centers are now in the eyes of at least one state actor, legitimate military targets. And if researchers like Zachary Kallenborn of King’s College London are to be believed, data centers will become increasingly targeted by both cyber and physical attacks as they become hubs for military information. These attacks are not limited to physical damage; beyond drone strikes, the conflict has also exposed how control over energy can be leveraged to pressure militaries that depend on AI capabilities for their operations.

The energy shock

AI data centers are power-hungry by nature, consuming roughly three to five times more electricity than conventional facilities. That appetite was already straining grids and driving up costs before the conflict. In the United States, residential electricity prices had risen nearly 30% since 2021. According to a Bloomberg analysis, electricity prices in regions with heavy data center activity are now up to 267% higher than they were five years ago. Notably, none of these projections accounted for the possibility of war disrupting global energy markets.

Now, as the conflict in the Middle East continues to disrupt the flow of crude oil through the Strait of Hormuz, the narrow passage through which roughly 20% of the world’s daily oil supply passes, energy prices are expected to remain volatile.

For data center operators, this means the baseline cost of running their facilities just jumped, and the timing could not be worse. Just days before the strikes, seven major U.S. tech companies signed the White House Ratepayer Protection Pledge, committing Amazon, Google, Meta, Microsoft, OpenAI, Oracle, and xAI to build or buy their own power generation for data centers and to pay for all necessary grid upgrades, whether or not they end up using the electricity.

It was a political gesture designed to shield American households from bearing the cost of the AI boom. But in a world where a war in the Middle East sets energy prices, that pledge becomes an open-ended financial commitment made at precisely the wrong moment.

For hyperscalers, a prolonged conflict means they will pay more to keep the lights on, which could lead them to build fewer data centers, build them more slowly, or build them elsewhere.

And the ‘somewhere else’ question is precisely where the conflict’s longer-term consequences begin to take shape.

The wiring problem

The Gulf’s appeal to Big Tech rested on a combination of cheap energy, sovereign capital, and a sales pitch about stability. As of now, the first two continue to stand; however, the third has taken a serious hit, and the damage goes deeper than the physical strikes.

Currently, 17 submarine cables pass through the Red Sea, carrying much of the data traffic between Europe, Asia, and Africa, while additional cables run through the Strait of Hormuz to serve countries including Iran, Iraq, Kuwait, Bahrain, and Qatar. With both passages now effectively closed to commercial traffic simultaneously, the region’s digital connectivity faces an unprecedented disruption.

As Doug Madory, director of internet analysis at Kentik, told Rest of World, closing both choke points simultaneously would be globally disruptive and has likely never happened before.

Then there is the ability of the conflict to force regional powers to contemplate that they can continue financing the AI buildout.

The funding freeze

Saudi Arabia, the UAE, Kuwait, and Qatar collectively manage some of the world’s largest sovereign wealth funds, which, following Trump’s 2025 visit to the region, pledged hundreds of billions of dollars in U.S. investments, much of it directed toward AI infrastructure, data centers, and projects like Stargate.

According to The Financial Times, three of the four major Gulf economies have begun internal reviews of those commitments, examining whether force majeure clauses (contract provisions that allow parties to exit agreements during extraordinary circumstances) could be invoked.

The Information separately reported that the war is putting roughly $300B in Gulf AI spending at risk.

As the conflict drags on, the financial pressure on Gulf countries continues to mount, with disruptions in tourism, aviation, and defense expenditure forcing them to review their long-term capital commitments.

This matters because the AI buildout has been structured around a specific financing model. Mega-projects like Stargate depend on a mix of corporate capital, debt financing, and sovereign co-investment.

If Gulf capital slows, delays, or retreats, the ripple effects will be felt not just in the region but in Silicon Valley venture rounds, U.S. infrastructure projects, and the private credit markets that have been eagerly lending against AI-linked assets.

Pricing in conflict

While none of this means the conflict will stop the AI buildout, it does mean the economics of where and how that compute is built are changing in ways that will outlast the conflict.

The economics of AI, which until now factored in model performance, latency, and cost per query, will have to include a variable that was not on any vendor’s slide deck six weeks ago: the probability that a regional conflict can take data centers offline, spike energy bills, or sever the cable carrying the data.

The AI industry built its infrastructure plans on the assumption that its biggest constraints would be technical. The war in Iran is a reminder that the oldest form of disruption is still a bomb, and the most stubborn bottleneck is still the price of keeping the lights on.