It pains me that I have to write this, but Gaia-X is a harmful and expensive distraction, and it is not doing anything that will ever get us a “European cloud”, not even indirectly. Its very existence is holding back progress. For this reason, Gaia-X should be abandoned, and we should try to learn as much as possible from its failure, so we can try something else. I provide some inspiration at the end of this post.

As one can imagine, the Gaia-X organization does not agree with my thesis, and in response to this post they specifically state that they are on the path of implementing the original Gaia-X plan, which was never “An EU Cloud”. They also are proud of their work on interoperable Data Spaces and open source software they have authored along the way. I urge you to read their full response further down this page, and I thank the Gaia-X people for being so open and approachable.

I’ve previously written extensively on what we could do in Europe to improve the cloud situation, so I do not come at this from an anti-European or pro-Big tech perspective. Far from it in fact. While this article is painful, I do feel it is necessary.

Europe currently lacks credible full service “hyperscalers”. I’ve only spotted two European companies that come close to credibly offering a basic set of cloud native services (Exoscale and Upcloud), and neither is even of “megascaler” size. Perhaps kiloscaler.

Whenever this gets discussed as a problem, people wheel out Gaia-X as “Europe’s answer to Amazon Web Services”. In fairness, neither the EU nor Gaia-X claims that this is what they are doing. It might surprise you, but Gaia-X is not even an EU project in any wayThe EU, somehow through Capgemini, has however supplied 1 million euros directly, and probably more indirectly. Most of the funding appears to come from European countries and industry. Gaia-X itself does not publish useful annual reports that I can find, although they did file 2022 fiscal numbers. ! However, that is what people often mistakenly assume Gaia-X is: a European Union initiative to build a cloud that can compete with AWS.

So what is Gaia-X then? From the website: “an initiative that develops, based on European values, a digital governance that can be applied to any existing cloud/ edge technology stack to obtain transparency, controllability, portability and interoperability across data and services”. If you know what this means you are smarter than I am.

Elsewhere we read “[The] Gaia-X Framework builds upon the evolution of the X-Model and enables the transition from disjoint data & infrastructure ecosystems, to composable, interoperable & portable cross-sector data sets and services”. Ok? And to clarify, “Our outcome will not be a cloud”. If anyone was confused on this, it is because they did not read the Gaia-X homepage.

An incorrect but popular interpretation is that Gaia-X is attempting to write up a set of cloud standards, and that existing (European) players could then get certified that they can deliver services based on those standards. This could then foster a broad market for such standardised services. Not the worst idea, but also very hard to do if you want this to be practically useful and competitive.

When we look into the Gaia-X standards written so far, we find a random collection of half-finished documents that do not talk about any of the must-have cloud-native operations like object storage ("S3"), virtual machines (“EC2”), basic web hosting, infrastructure as code, distributed (“Aurora/Spanner”) or even regular databases (“RDS”). We do get complicated definitions of “Ontologies for Data Exchange”, whatever those may be.

Furthermore, many key documents appear to not have been written yet, others stem from 2021 and haven’t been updated since. On top of all this we find a separate somewhat newer stack of specifications and software called the Gaia-X Digital Clearing House, but despite the name, it is not clear to me what this is.

So the problem is not only that the documents are not finished (although they aren’t). They are also the wrong documents! The set of Gaia-X deliverables contains no cloud service definitions at all – and this is by design.

The incredibly hard task of standardising rapidly developing commercial cloud services is not even in scope. This is why we only find complicated cloud meta-definitions and ontologies, and not something anyone could actually implement, let alone competitively.

The actual and official Gaia-X product is a set of standards so we can communicate about (as yet to be written) cloud/data standards. This is two steps removed from being useful - not a cloud, not a standard for a cloud, but a way to exchange authenticated data about standards compliance.

When asked, Gaia-X people will tell you it is nearly impossible to write actual standards for a cloud, since everyone wants something different. However, if this is all so hard, then why is it useful for Gaia-X to create solutions for talking about such impossible standards? I could not get an answer to this question.

There is however a weird catalogue of “Gaia-X compliant services” that lists things like “Private Cloud” hundreds of times - which is all not very credible, since these services do not appear to have an actual Gaia-X standard associated with them. The various services offered are “Gaia-X Compliance 22.10 offerings, completed with a CISPE Label extending the Gaia-X Label 22.11 criteria with sustainability criteria of the Gaia-X Label 23.11 (PRCD 23.11)”. I have tried very hard to figure out what this means. As far as I can tell these are statements that a “Private Cloud” gets a “Private Cloud” label, but that there is no actual definition what this then is.

It also appears that Gaia-X has many members that would not benefit from helping create new European competitors for themselves. Think of companies like Amazon, Huawei, Microsoft, Alibaba Cloud, Google and IBM. Perhaps having the competition on board is also behind the failure to get anywhere.

How Gaia-X harms European cloud initiatives

Gaia-X has attracted hundreds of other members, showing just how powerful its draw has been. And this is a big problem. Not only is Gaia-X very unlikely to be useful anytime soon, it is distracting people into thinking something useful is happening. Its mere existence is sucking up the oxygen and euros for initiatives that could be worthwhile.

As an example, I’ve often had to explain to policy makers and politicians that Gaia-X is not in fact what they were assuming it was (“EU AWS”). These people would then likely also have discounted actually productive initiatives since they thought Gaia-X was already on the path to creating a European cloud economy - which it very much is not. People also use Gaia-X as a weak excuse to do nothing, since Gaia-X “is already the official European initiative”.

In addition, it is not clear how much money has been spent on Gaia-X (related) projects, but from looking around a bit, at least the German government appears to be paying tens of millions of euros per year. I hear the sum might be over 200 million euros. We could do so many much more useful things for that kind of money!

So in the interest of getting actual credible cloud native services going in Europe, something we desperately need in these trying times, Gaia-X should be abandoned.

Not only is it using up lots of money and resources that could better be spent elsewhere, its very existence fools people into thinking something useful is happening.

With all appreciation for the hard work that has obviously gone into the many Gaia-X meetings and documents, now is the time to stop.

So what SHOULD we do?

A full post-mortem on Gaia-X would likely teach us a lot of things. One key thing appears to be that Gaia-X was very much a top-down affair. Large corporate entities became members, but did not involve anyone who ever personally rolled out cloud-based services. Or perhaps they did initially bring such people, but these then left the scene because nothing practicable was going on.

Any future initiatives should be self-hosting as quickly as possible, showing the value of standardised services by using them in production. This keeps everyone focused on the goal.

I’ve spoken before about the cloud in terms of being like IKEA (hard to compete with because it offers everything), and in terms of it being like Airbus (doing highly impressive technical things).

There have been calls for creating an “Airbus for Cloud”, but we should keep in mind how painful and slow the birth of Airbus was.

However, we might also recognize that many European places (like Leaseweb, Hetzner, OVH, Scaleway, team.blue, Ionos and others) together host astounding amounts of servers, spewing out many terabits/s of services. Their qualities are concentrated at the lower end of the hardware & service stack, but these are very important qualities. Also, these companies provide very affordable services, often much cheaper than their hyperscaler competitors.

If we could somehow marry the cloud native smarts of smaller cloud-native providers like Exoscale and Upcloud with these hosting powerhouses, we might relatively quickly at least do ‘somewhat of an Airbus’. If we work on practical standards and use those to cooperate, this is very feasible.

Such a combination is not an “Amazon-beater” out of the box, but it would at least provide a credible alternative to companies and governments that can think beyond buying from the dominant players, and that perhaps care about digital autonomy.

And if Europe wants to have any position in the cloud, this would be a great start.

Data Spaces

In this post I have mostly spoken about Gaia-X in the context of the cloud. However, Gaia-X has also been working hard on “Data Spaces”, and I think they find this currently to be more important than the cloud efforts.

“Data spaces are the digital representation of existing physical or natural or industrial or social ecosystems”, is what we read on the website. As with the cloud work, the Data Spaces effort is not delivering an implementable service. It does however create verbiage through which such a service could eventually be built.

Key to the Data Spaces idea is that data is very well labeled, and could be shared while retaining metadata on who owns the data, and under what conditions it could be shared further. There are also semantics on how to revoke data. Data also comes with traceability of consent of the sharing. Whenever possible, “the data are not stored centrally, but rather at the source. Thus, they are only transferred through semantic interoperability as necessary”.

In some way this work is more exciting than attempting to standardise standards for talking about cloud standards. At least this stuff is rather novel, and indeed solving a problem not too many other initiatives are working on. However, as with the cloud work, it is all rather abstract, with actual implementations considered to be out of scope, making it all somewhat theoretical.

Gaia-X Response

I’m grateful for the Gaia-X organisation for the productive conversations we’ve had, and they have provided this statement as a response to this post:

Attached: TOGETHER TOWARDS A FEDERATED & SECURE DATA INFRASTRUCTURE - Gaia-X 2024

Some further links

• Inside Gaia-X: How chaos and infighting are killing Europe’s grand cloud project, 2021 piece
• European cloud project Gaia-X is stuck in the concept stage, 2022, “The project is so far more an idea than a concrete plan”
• Scaleway will not renew its Gaia-X membership., 2022, “Accepting all the non-European dominating cloud service providers without any limitation had a significant impact: once they joined the technical committees, these dominant entities and their ’tech diplomats’ flooded other contributors with orientations, requirement proposals and comments that the European collective could not possibly cope with”
• Gaia-X project doesn’t have a future, claims Nextcloud boss, 2024, “Just a bunch of standards and certifications, which are not really helpful, unfortunately”
• Gaia-X: Has Europe’s grand digital infrastructure project hit the buffers?, 2024, “Talk of data spaces and digital clearing houses is far removed from the virtual hyperscaler plan, but was always part of the initiative’s ambition” - I’m still not sure what digital clearing houses are, though.

– bert hubert (bert@hubertnet.nl)