Author: David, Deep Tide TechFlow

On March 1, Iranian missiles and drones struck the Gulf region, with one landing on an Amazon data center in the UAE.

The data center caught fire, lost power, and about 60 cloud services were interrupted.

One of the largest AI users globally, Claude, runs on Amazon's cloud. On the same day, Claude went down worldwide.

Anthropic's official statement was that user surges overwhelmed the servers.

At the time of publication, there were still complaints on social media about Claude's service being unavailable; on the well-known prediction market Polymarket, a topic predicting "Claude will go down several times in March" has emerged.

If it is ultimately confirmed that Iran was responsible, this will be the first time in human history:

a commercial data center has been physically destroyed in war.

However, why would a civilian data center be bombed?

Two days earlier, on February 28, the United States and Israel conducted a joint airstrike on Iran, killing Supreme Leader Khamenei and several high-ranking officials.

The intelligence analysis, target identification, and battlefield simulation for this air strike were significantly assisted by Claude. Through collaboration with the military and data analytics company Palantir, Claude had long been embedded in the US military's intelligence system.

Ironically, just hours before the airstrike, Trump ordered a complete ban on Anthropic because Anthropic refused to hand over AI without restrictions to the Pentagon. But bans aside, the war still had to be fought.

To remove Claude from the military system, the official statement indicated it would take at least six months.

So, with the ban barely dry, the US military took Claude to bomb Iran. Then Iran retaliated, and missiles fell on the data center running Claude AI.

Image source: Bloomberg

The data center was likely not targeted but caught in the crossfire. However, whether the missile was aimed at the data center or not, one thing is certain:

The truth is within artillery range, and AI is also within artillery range.Both sides, the one firing the artillery and the one being hit by it, are.

AI infrastructure built on a powder keg in the Middle East

In the past three years, Silicon Valley has moved half of the AI industry to the Gulf region.

The reason is simple. The UAE and Saudi Arabia have the world's wealthiest sovereign funds, cheap electricity, and a rule:

If you want to serve my customers, the data must exist on my territory.

Thus, Amazon opened data centers in the UAE and Bahrain and has invested $5.3 billion to open another in Saudi Arabia; Microsoft has nodes in both the UAE and Qatar, and construction in Saudi Arabia has been completed.

OpenAI, in partnership with NVIDIA and SoftBank, is building an AI park in the UAE worth over $30 billion, claimed to be the largest computing power base outside the US.

In January this year, the United States just signed an agreement called "Pax Silica" with the UAE and Qatar. Translated, it means "peace of silicon," which sounds beautiful.

The core content of the agreement is to control the flow of chips, ensuring advanced chips do not fall into Chinese hands.

In exchange, the UAE secured licenses to import hundreds of thousands of NVIDIA's most advanced processors each year. G42 in Abu Dhabi severed ties with Huawei, and Saudi AI companies promised not to purchase Huawei equipment...

The entire Gulf region's AI infrastructure, from chips to data centers to models, has shifted fully towards the United States.

These agreements consider everything from chip export controls, data sovereignty, investment reciprocity, to technology leakage risks.

But not one of these provisions considers that someone might use missiles to bomb data centers.

An international security scholar from Qatar University remarked on the fire at the Amazon data center that I found particularly fitting:

“These security frameworks are designed for supply chain control and political alignment, while physical security has never been on the agenda.”

The story of cloud computing told for a decade is about resilience, redundancy, and decentralization. But data centers are buildings with addresses, with walls, roofs, and coordinates. No matter how advanced your chips are, if the data center is bombed, it is bombed.

"Cloud" is a metaphor, but data centers are not.

AI appears intangible, running in code, floating in the cloud. But the code runs on chips, the chips are housed in data centers, and the data centers are built on Earth.

Who will protect AI?

Amazon's data center this time could be considered collateral damage, or optimistically, mis-targeted.

But what about next time?

In a context of increasing global geopolitical conflicts, if there is an AI model running in your data center that assists your opponent in target identification, your opponent has every reason to treat your data center as a military facility.

This question has no answer in international law either.

Current war laws address "dual-use facilities," but those provisions refer to factories and bridges; no one has considered data centers.

A data center that runs bank transactions during the day and helps the military with intelligence analysis at night—does it count as civilian or military?

In peacetime, data center site selections consider latency, electricity rates, and policy incentives... When war comes, all of this becomes irrelevant; what matters is how far your data center is from the nearest military base.

Therefore, this bombing has begun to shift everyone's attention.

Previously, everyone was discussing the same anxiety: would AI replace my job; but no one discussed another question:

How vulnerable is AI before it replaces you?

A regional conflict brought down the Middle East node of the world's largest cloud service provider for an entire day; and this was just one data center.

Currently, there are nearly 1,300 hyperscale data centers worldwide, with 770 more under construction. These data centers consume increasing amounts of electricity, water, and money, while also carrying more and more—your deposits, your medical records, your takeout orders, even military intelligence of a certain country...

Yet, the plans to protect these data centers remain, to this day, possibly just fire protection systems and backup generators.

When AI becomes a nation’s infrastructure, its security is no longer just a company’s issue. Who will protect AI? Cloud vendors? The US Pentagon? Or the UAE's air defense system?

This question was theoretical just three days ago. Now it is not.

AI is within artillery range. In fact, it is not just AI. In this era, what is not within artillery range?