AWS Will Run Its European Sovereign Cloud Without the Global Backbone for Several Hours

On Saturday, October 24, AWS will cut its European Sovereign Cloud off from the AWS Global Network backbone for several hours. The region is expected to keep running. What that proves is a narrower thing than the argument now going on around it, and AWS says so itself.

Two systems come out. One is the Global Network backbone, the private network that moves operational data between AWS locations. The other is a dedicated channel for limited data transfers, switched off entirely. Traffic then falls back to the public internet over the region’s own links from European ISPs. Stéphane Israël, managing director of the AWS European Sovereign Cloud, says a team "composed entirely of EU residents working in the EU" will run it on the region’s own hardware and software, out of data centers in Brandenburg that hold a local in-EU copy of the source code. Customers may see short disruptions while routes converge at either end.

The caveat comes from AWS. Neither system is operation-critical. That is why the region survives without them. It is also why the region uses both every other day of the year. Data gets a sharper line. Customer content and customer-created metadata stay in the EU. Some operational data leaves. Israël writes that the exercise is built to line up with the EU Cloud Sovereignty Framework and Germany’s BSI C3A criteria, and that customers can put the results into their compliance packages. Regulators hear about it afterwards too. The post stops well short of claiming certification. It does not pretend the region runs isolated the rest of the year either.

None of that care has settled anything. Fadi Dasus, a senior cloud security architect, split the question in two:

Sovereignty on a US-owned cloud in the EU comes down to two areas. The technical one: does the service stop if the link to the US is cut? The legal one: can US law order it to hand over your data, or to stop serving you? If the test goes as planned, the first area is no longer entirely in American hands. The second still runs through the US parent. So, this is real progress for resilience. It just isn’t sovereignty.

Lars Larsson Lundsten, field CTO at Elastisys, which sells a European alternative to US cloud platforms, put it harder. "You can’t solve legal problems with technical solutions" he wrote, while calling the architecture a marvel of engineering. He also credited AWS with restraint, noting it is "extremely careful not to claim anything besides what this is able to actually prove."

The legal framing drew its own pushback. Commenter Roelf Zomerman argued that the CLOUD Act and the EU e-Evidence regulation both run through criminal investigations, with judges, provider evaluation and data limits attached, rather than working as a general power to demand data. Building your own, he added, or hosting with EU providers, "doesn’t automatically imply you are safe from these laws." Anders Bjørnestad, a principal consultant and AWS Community Hero, corrected a related assumption:

There are two aspects of this discussions. "Can someone get access to my data", and "can someone shut me off". What is tested here is the last part. The first part can be solved with encryption if you do it the right way. And all companies doing business in/with the US in some form are bound by US part of law, it is not only ownership that counts.

The hardest technical question came from a municipal IT manager rather than a vendor. Odd-Arne Haraldsen, who runs IT operations at Svenljunga Kommun in Sweden, called the design impressive, then asked what a few hours actually establishes:

How long can it actually operate independently? How long before systems fall out of sync? How long before authentication tokens, certificates, or licences expire? And how long before dependencies on external control planes begin to undermine that resilience? Redundancy is not the same as independence, and continued operation is not the same as sustained autonomy.

So the overclaiming charge misses. AWS has not stretched its evidence. The gap sits between what the exercise measures and what buyers keep asking. Take the obvious case. The post never says what happens to global services while the backbone is down. One commenter asked how IAM works when the region is cut off from the rest of AWS. Another said he could not picture DNS and internet routing holding up. The announcement addresses neither. Commenter Jarrod B. went at the audience question instead: the exercise "isn’t meant to appease legal purists; it’s designed to satisfy European risk officers and regulators." AWS hands the results to compliance teams, so he may be reading it right.

One exchange put the engineering in longer perspective. A commenter recalled a global AWS outage that appeared to show parts of AWS depending on us-east-1 and unable to run anywhere else. Lundsten gave ground in reply: if AWS has engineered that dependency away, its architecture is better for it.

AWS has given no launch date for the region. It calls the thing "a new, independent cloud for Europe." An update follows October 24.