TL;DR

• Cloud providers own hardware, but make most money on a unified software stack.
• What if you could build such a unified software stack that works on top of anyone else's data center?
• A cloud provider using this, opens up a unique offering of freedom in physical placement of storage and compute.
• It would be extremely hard to bring this to market, for many known reasons, and probably many unknown reasons.

The Cheap Idea: A Sovereign Virtual Cloud Provider

Hosting your business critical services, and PII-laden datasets with a sovereign provider is a hot topic. So hot, that AWS has dedicated a complete section of their compliance marketing to it, including videos showcasing extremely European things, like bicycles, high speed trains, well maintained highways, and — presumably universal — health care. They use words or phrases like autonomy, ownership and digital sovereignty. Unfortunately, none of that matters, because Amazon is a US based company. And US companies comply with US law enforcement before anything else. This was never a concern, until doubt was cast on the US' reliability as an ally.

All this created a world where sovereign is no longer a marketing term, but has come to mean very concretely: absolutely not linked to any nation outside EU, and preferably not closer to Russia than Germany. This new reality comes with obvious business opportunies. One of them being a viable cloud platform that lives in EU, under EU business ownership, and complying only with EU law.

Building such a business is time and captial intensive. Even US companies take years to obtain permits and other local approval for building data centers in Europe. It seems unlikely that us Europeans will move more quickly. But, what if you could build a virtual layer on top of existing hardware?

What Is a Cloud Provider?

A common misconception is that cloud providers make most of their money charging a premium on renting computers by the hour. While this is certainly a part of the business, by far the largest margin of a cloud provider is made on the software stack that allows for selling a homogeneous set of hardware disguised as a diverse, and specialised set of services. You can get this confirmed by any business that has attempted to compete in cloud computing with a bare metal product.

In 2025, AWS has over 240 individual services in their catalogue (according to Google AI). Quite obviously, they do not operate equally many different hardware profiles and sets of networking equipment in their data centers each tailored to a single service. And, perhaps less conspicuously, all of these services are priced according to customers' willingness to pay and perceived value; not according to hardware resource usage. The higher up the stack, the bigger the margin. Cloud providers are much more interested in managing your email servers, databases, and message queues, than selling bare CPU cycles or virtual disks.

Once you realise the better part of the margin is on software, not data centers, the question is: can you build a cloud provider by just building the software layer while leveraging existing (European) data centers? Then you could share the margin with data center providers, who in turn additionally benefit from the scale of a single customer base, while they can continue competing on their existing unique selling points (locality, network uplink speeds, cost, reliability, etc.).

How Would This Work?

You would offer an initially limited, but nonetheless useful set of cloud services through a unified API for managing resources like any other provider. The one key differentiator is: customers have full control over who owns and operates the physical layer of the services. This is something that inherent to their business model, scale, and ownership structure, a tier 1 provider (AWS, GCP, MSFT) could never offer. The premise of the business viability is whether you can bring the price down to a level where the premium over an existing tier 1 provider is small enough that customers are prepared to pay it for guarantueed, true sovereignty as opposed to AWS' marketing nonsense showing German trains.

Why Is It Hard?

In no particular order, here is a very far from exhaustive list of reasons why it is extremely hard to do this:

• You have to do DevRel exactly right to attract initial users.
• You have to get the initial service mix exactly right to attract initial customers.
• You have to launch with an impressive set of data center partners with plenty of availability to scale if necessary.
• You have to convince all data center partners to put in a probably less than trivial effort to integrate with your cloud management APIs.
• You have to build a unified software defined networking layer covering heterogeneous data centers, with different points of presence, networking hardware, differently managed IP spaces, and varying degrees of BGP maturity. You will probably end up installing your own networking equipment co-located with some or many of the data center providers.
• You have to recruit a handful of very expensive key product and engineering people away from an existing tier 1 cloud provider to avoid some more subtle pitfalls.
• You will have to direct extreme focus to a "just works" experience; don't leave customers with that OpenStack-feeling.
• Etc.

The more you think about it, the longer this list gets. But on the positive side: you don't have to build any data centers.

And one more thing is easy as well. You will likely beat any European Union initiative to market.

Why Don't I Do This?

I estimate the required seed funding to get this idea into a shape viable for a handful of cherry picked launching customers to be on the order of 2M EUR. I don't have that. And the next round will have to be an order of magnitude on top in order to scale to many data center partners, whom initially will have to be incentivised.