Physical risk is intruding on the AI infrastructure boom
The push to build massive new data-center capacity in Gulf states is running into a harder constraint than power, land, or chips: war risk.
Ars Technica reports that Pure Data Centre Group, a London-based developer operating or developing more than 1 gigawatt of capacity across Europe, the Middle East, and Asia, has paused all Middle East project investments after one of its facilities was damaged by an Iranian missile or drone attack. The decision reflects a problem that Silicon Valley’s infrastructure ambitions cannot easily abstract away. Data centers may be digital businesses, but they remain physical assets exposed to geography, insurance markets, and military conflict.
That exposure is especially relevant now because Gulf states have become central to long-range plans for AI and cloud expansion. Cheap energy, political willingness to fund large projects, and ambitions to become regional compute hubs have made the region attractive. But the same concentration strategy looks different when war damage becomes uninsurable or recurrent.
What happened
According to the report, the Iran war began on February 28 with a US-Israeli attack on Iran, followed by Iranian retaliation that targeted shipping, US military bases, energy infrastructure, and data-center facilities in the Gulf. Iran reportedly struck two Amazon Web Services data centers in the United Arab Emirates directly, while a near miss from a one-way attack drone damaged a third AWS facility in Bahrain.
AWS said the attacks caused structural damage, disrupted power delivery, and triggered fire-suppression systems that led to water damage. The incident rippled into widespread service disruptions affecting customers including banks, payment platforms, the ride-hailing company Careem, and Snowflake.
Amazon also reportedly waived customer charges in its Middle East cloud region for all of March 2026, a reminder that infrastructure damage does not stop at repairs. It can cascade into service credits, customer distrust, and delayed enterprise adoption.
Why the pause matters
Pure DC chief executive Gary Wojtaszek told CNBC that no one would commit new capital at scale until conditions settled down. That remark is significant not because one company is taking a temporary pause, but because it captures a likely financing logic across the sector.
Large data centers require heavy upfront investment and long payback periods. Their economics depend on confidence that operations can remain stable long enough to monetize multi-year customer contracts. If assets are exposed to attacks that insurers will not fully cover, then the business case changes quickly.
In that environment, companies may still maintain existing operations, but new projects become harder to justify. Investors will demand higher returns, more political assurances, or alternative locations. Some projects may slow. Others may move.
AI scale meets geopolitical reality
The timing is awkward for the industry because the Middle East has been framed as a strategic frontier for AI infrastructure. Gulf governments and global tech companies have discussed enormous plans for compute, cloud services, and regional digital sovereignty. Those ambitions often assume that capital can be deployed once energy and policy align.
The current disruption shows a missing variable in that model. Physical resilience and conflict exposure now belong alongside power procurement and accelerator supply in the planning stack. That is a meaningful shift, especially for AI infrastructure, which is already expensive, power-hungry, and politically sensitive.
A damaged logistics warehouse is costly. A damaged cloud region can be more destabilizing because so many downstream services depend on it. When the customer base includes banks, transport platforms, and enterprise software providers, data-center reliability becomes part of wider economic continuity.
Insurance may be the hidden bottleneck
The report emphasizes uninsurable war damage, and that may turn out to be the key issue. Tech companies can sometimes price in higher construction costs or redundancy requirements. What is much harder to absorb is the prospect that a facility in an active conflict zone cannot be adequately insured at commercially reasonable terms.
Insurance is often treated as back-office plumbing until it fails. But for capital-intensive infrastructure, it is a gating mechanism. If lenders, operators, or hyperscale customers cannot get comfortable with the risk profile, projects stall regardless of how attractive the market looks on paper.
That could reshape where the next wave of AI buildout lands. Rather than concentrating new capacity in geopolitically exposed hubs, firms may favor locations with greater physical security even if they are less efficient on energy or tax terms.
What this means for the industry
The broader lesson is that AI infrastructure strategy is becoming inseparable from geopolitical risk management. Companies are not merely deciding where electricity is cheap or regulation is favorable. They are deciding where high-value compute clusters can survive shocks without collapsing service promises or balance-sheet assumptions.
That recalculation may not end Gulf ambitions, but it could alter them. More redundancy, harder security requirements, higher financing costs, and longer decision cycles are all plausible outcomes. Some companies may continue to build, but with revised assumptions about resilience and exposure.
For now, the industry has a clear warning. The global race to deploy AI capacity is often described as a contest over chips, talent, and energy. It is also a contest over stability. In the Middle East, the latest attacks show how quickly data centers can move from being symbols of digital modernity to front-line strategic liabilities.
This article is based on reporting by Ars Technica. Read the original article.
Originally published on arstechnica.com
