About Those Data Centers. Yeah, the Ones with Targets on Satellite Imagery. Those Data Centers
May 4, 2026
Another dinobaby post. No AI unless it is an image. This dinobaby is not Grandma Moses, just Grandpa Arnold.
I read “Amazon Stuck with Months of Repairs After Drone Strikes on Data Centers.” The write up contained a number of factoids I found interesting. One example is the statement “AWS stops billing Middle East cloud customers as repairs to war damage drag on.”
Let’s think about this statement, assuming that it indeed accurate.

Thanks, Venice.ai. Good enough for an act of generosity.
Amazon, for some in the conflict zone and peripheral areas, synonymous with cloud computing. I would argue that in certain areas Microsoft Azure has a more prominent profile. For the present discussion, let’s focus on Amazon. The company does not publish a name, location, and facility capability description for each of its infrastructure locations in Bahrain, UAE, Saudi Arabia,Oman, and Qatar. But Amazon lists AWS regions and availability zones and direct connect locations, usually providing the city and the country. With that information, it is not too difficult to pinpoint the exact location of Amazon facilities by [a] asking a taxi driver if one is in country, [b] making some contact with government-linked business “facilitator,” or [c] getting some gig work market research via Fiverr.com.
My point is that the article makes clear that at this point in time (May 2, 2026 at 236 pm US Eastern time) Amazon has a number of facilities that could become targets at any time. This adds to the challenge Amazon and other data center operators face in contentious parts of the world. How can these facilities defend themselves against drone attacks? If attacked, how can the damaged facility be repaired and returned to “normal” operation?
The reason I bring this up is that the cited article makes a big deal about Amazon not billing Middle East cloud customers for the damage that has occurred. That’s okay. There are some problems that plague the approach:
- So far Amazon has been able to cope with the damage to the handful of its data centers; however, the implication in the write up in my opinion is that getting these facilities back to pre-attack status is not quick and easy. Those two factors translate into costs. Amazon is investing in AI; the company is terminating some employees who are not productive. The “grace period” cannot be extended indefinitely it seems to me.
- If hostilities flare up, data centers are easily identified without help from nations with satellite imagery. This means that a coordinated attack could take out a number of data centers simultaneously. I can envision war planners thinking, “Why not take out the Bahrain and Saudi facilities on one day and then go after the Microsoft Azure facilities the next day?”
- Companies providing services to the large data center operators can also be easy and vulnerable targets. What defensive measures does Equinix-type companies have in place?
Let’s think about how commercial firms like Amazon can protect these quite large, easily findable, and highly vulnerable data centers. Options range from the use of World War II methods like sandbags, temporary defense shielding, and Kevlar shrouds. These are cheap and visible “protections.” The problem is that a drone only has to fly into a power or cooling vent, and the “off switch” can be flipped.
More robust defensive measures run into the same problems as the repairs to damaged facilities: Time, resources, and costs. Data centers vulnerable today will be vulnerable to some indeterminate time in the future. Thus, until the facilities are hardened, they are at risk.
If we think in terms of an aggressor attacking data centers owned by US companies in other countries, the attack surface becomes much larger.
Several observations are warranted:
- The success of attacks on the Amazon facilities makes clear that these structures were not designed, engineered, and constructed to withstand a drone- or other kinetic attack. That seems to have been what I would call an ill-considered decision. The reasons may range from keeping costs down or MBAs in the US don’t think too much about asymmetric issues.
- Data centers now under construction may require expensive hardening if these facilities are located in regions where they become large, easily located targets. Stated another way: The cost of those under construction data centers are likely to go up and require more time before they go online and produce revenue for their operators.
- Future data centers will warrant some design, engineering, and construction scrutiny. The success of Ukraine drone in damaging distant Russian infrastructure is a new factor at least for the “Z” folks.
The Ars Technica write up includes this passage:
The latest AWS status update comes just after another data center developer, the London-based Pure Data Centre Group, said it will pause Middle East data center investments until the ongoing Middle East conflict subsides.
Pure Data is thinking clearly. However, can one really go home again or step into the same river twice? The massive US investment in data centers makes these facilities even more attractive targets than they were when Amazon experienced its first attacks. Going forward, addressing the issue means data centers are going to get more expensive, not less expensive. Statements like “stop billing” might be tough to honor.
Stephen E Arnold, May 4, 2026
Comments
Got something to say?

