Artificial Intelligence

Lights Out for Arab AI

Lights Out for Arab AI

On March 2, 2026, the world witnessed a significant disruption in artificial intelligence (AI) services, particularly affecting major platforms like Claude, ChatGPT, and Google Gemini. This incident coincided with Iranian retaliatory strikes on Gulf infrastructure, raising questions about the vulnerabilities inherent in the electrical infrastructure that supports these AI systems.

The Outages

Early on the morning of March 2, users of Anthropic’s Claude AI reported widespread outages, with nearly 2,000 outage reports logged on Downdetector at the peak of the disruption. The web interface, mobile applications, and login systems were non-functional for over four hours. OpenAI’s ChatGPT faced a similar fate, with an ongoing disruption lasting 14 hours and over 20,000 users reporting issues. Google’s Gemini API also logged 27 outage reports within the same 24-hour period.

While Anthropic attributed the outages to “unprecedented demand” following a political incident involving the Pentagon, the timing of these disruptions raises concerns about their connection to the Iranian strikes on Gulf infrastructure. This incident highlights a critical vulnerability: AI does not operate in a vacuum; it relies heavily on electrical infrastructure.

The Importance of Electrical Infrastructure

Artificial intelligence systems require enormous, continuous, and uninterruptible quantities of electricity. A single hyperscale data center can consume between 50 to 100 megawatts of power—enough to supply electricity to a small city. The GPU clusters that train cutting-edge AI models require months of continuous power. If power is lost during a training run, it can result in the loss of weeks of computational work that cannot be recovered.

This electricity is delivered through the power grid, which relies on transformers. High-voltage power transformers, particularly the 400kV units that step down transmission power for data centers, are custom-built for each installation. These transformers weigh between 100 to 400 tons and have manufacturing lead times of 18 months to three years. Currently, there is a global shortage of these critical components, which are often left undefended in open yards behind chain-link fencing, connected to overhead transmission lines.

The Ukraine Lesson

The vulnerabilities of electrical infrastructure have been starkly illustrated by the ongoing conflict in Ukraine. In 2025, Russia launched a series of attacks targeting Ukrainian rail infrastructure, focusing on power substations rather than the trains or tracks themselves. This strategy effectively crippled Ukraine’s rail network, which relies heavily on electric locomotives. Similarly, Ukrainian forces targeted Russian power transformers to disrupt their rail operations.

Both sides of the conflict have recognized that destroying transformers can lead to a systemic collapse. While tracks can be repaired quickly, the custom nature of high-voltage transformers means they cannot be easily replaced. The global supply chain struggles to keep up with the demand for these components, as evidenced by Ukraine’s difficulties in replacing its damaged transformers despite extensive support from Western allies.

The Events of March 1, 2026

On March 1, 2026, Amazon Web Services (AWS) reported that one of its Availability Zones in the UAE was impacted by objects that struck the data center, resulting in sparks and fire. Although AWS did not confirm whether this incident was linked to Iranian strikes, the fire department’s response involved shutting off power to the entire facility, including backup generators. This led to a significant outage affecting multiple AWS services, including EC2, Lambda, and S3.

The nature of the incident—described as an electrical event rather than a structural one—suggests that the data center’s infrastructure may not have been built to withstand such attacks. The fact that the entire building lost power due to the fire department’s protocol indicates a lack of hardened infrastructure standards, which are crucial for critical facilities.

Implications for AI in the Gulf

The implications of these events for the burgeoning AI industry in the Arabian Peninsula are staggering. With tens of billions of dollars being invested in AI compute infrastructure, the vulnerability of electrical systems poses a significant risk. As the Gulf region continues to develop its technological capabilities, the lessons learned from the ongoing conflicts and the recent outages will be crucial in shaping future investments and infrastructure designs.

Frequently Asked Questions

What caused the outages in AI platforms on March 2, 2026?

The outages were linked to simultaneous service disruptions in major AI platforms, including Claude, ChatGPT, and Google Gemini, coinciding with Iranian strikes on Gulf infrastructure, which raised concerns about the vulnerabilities of electrical systems supporting these services.

How does electrical infrastructure impact AI operations?

AI operations require vast amounts of continuous electricity. Any interruption can lead to significant losses in computational work, as AI models depend on uninterrupted power supply for training and operation.

What lessons can be learned from the Ukraine conflict regarding infrastructure?

The Ukraine conflict has highlighted the critical vulnerability of electrical infrastructure. Targeting power transformers can lead to systemic failures in transportation and other essential services, demonstrating the importance of protecting such infrastructure in any conflict.

Note: This article reflects the current understanding of the events surrounding the outages and their implications for AI infrastructure in the Gulf region.

Disclaimer: eDevelop provides blog and information for general awareness purposes only. While we strive for accuracy, we do not guarantee the completeness or reliability of any content. Opinions expressed are those of the authors and not necessarily of eDevelop. We are not liable for any actions taken based on the information published. Content may be updated or changed without prior notice.