Is sovereign AI dead? If so, is it a non-issue, or is it because accomplishing that goal has proven beyond the proponents’ ability? An EU initiative to build a sovereign cloud seems to have collapsed, but there’s still a lot of discussion on whether the EU needs one, and perhaps needs to take steps to reduce “technology dependence” on the US, so let’s look at it and try to see what’s wrong, and how it might be righted.
The initiative in Europe, Gaia-X, had from the first too many goals and too many masters. To some, the goal was to make the EU less dependent on US technology. To some, it was to ensure that EU companies were subject to EU law and regulations, for others it was simply a desire to stop paying US providers, and to some it was almost a social goal, to foster a bigger place on the world stage. In parallel with this, Gaia-X was a bit of a pawn in the differing approaches of France and Germany, the former seeing itself as a kind of protector of the old Europe and the latter seeing itself as an up-and-coming advocate of a rather different “new Europe” vision.
All of this churning situation plays out within a truth that’s often cynically stated as “The IQ of any group is equal to that of the dumbest person, divided by the number of people”. A government, of course, is the ultimate group. Standards processes are glacial because they involve a lot of conservative people, and governments have more conservatism and more people. Do the math, and take your time because they’re not going anywhere. And in any case, the goal may not be achievable, no matter how smartly you approach it.
Fundamentally, Gaia-X is a kind of industry group or standards body, and while I’m not a member, I talk to a few whose organizations are, and what I hear from them is familiar to anyone who’s participated in any consensus tech activity here in the US. There’s goal conflict, competition, ego, fear, and the ever-present debate on whether, if doing nothing is a risk, doing anything at all is more so.
Why did clouds take off in the US? Why did the Internet take off here, despite the fact that the seminal technology (the worldwide web, HTTP and HTML) really originated in the UK? While the EU is a capitalist market, few would disagree that the US is “more capitalist” than other markets. Companies are more risk-tolerant, more fast-fail, in the US. The venture capital markets here will take a risk on a lot of companies, expecting most will fail but a couple will pay back enough to make up for it. That thinking is less common in Europe. All this means that innovation doesn’t require consensus, and in fact it arguably shuns even the attempt at it. Nobody, in a society/economy built on taking advantage of opportunity, wants shared opportunities. Instead, the universal fear of first-mover advantage going to some other mover, drives everyone to try to move first and thus guarantees motion.
This difference in the mechanisms for and financing of startups is critical, and it is likely a reflection of the overall attitude of a society, an economy, toward innovation and change. If that’s the case, then there are underlying socio-economic factors that make the creation of something like Gaia-X difficult. To overcome this might require either subsidization by the EU, or higher prices for cloud users. For sure, it’s going to create a crunch in competing with long-established US firms on cloud feature development. Failure of the venture would risk putting EU companies who used Gaia-X at a competitive disadvantage.
Of course, many in the EU (governments, businesses, and consumers) think that EU companies are already at a disadvantage for lack of sovereign cloud. There’s the undeniable issue of being disadvantaged in tech skills by having something as fundamental as cloud services created offshore. There’s the question, recently raised, about the extent to which the US can be depended upon as it has been.
The admitted tech war between the US and China exacerbates the EU challenges, too, as the German Chancellor recently noted. Yet many of my friends in Europe have noted that they fear having to take sides in this, some note that they have done that already with the US pressure on Huawei, and all think that a two-country race means two-country progress on technical matters that would only put the EU further behind.
The problem is how to deal with this, and you probably already see that there are so many layers involved here (government, economic, social, technical, nationalistic) and so many views on each layer that even proposing a “solution” is almost presumptuous. One of my French friends quipped “Of course I have an answer. Of course everyone does, and they’re all different, so in the end there’s no answer at all.”
One answer long proposed is to “on-shore” that which has been off-shored, in compute and data. That means encouraging or mandating (directly or via sovereignty regulations) US cloud companies to host EU applications on facilities they deploy in the EU. That satisfies many, but not all. What I hear from Germany and the Netherlands is that this is a positive step; France and Belgium aren’t so sure. It satisfies concerns about snooping, particularly government snooping, but only if you believe a US company wouldn’t bow to government pressure here if it were applied.
Wall Street, using the term to refer as most do to the global finance market community, has its own concerns, particularly because cloud issues inevitably raise AI issues. One easy issue to understand is “competitive overbuild”. Can everyone build an Internet, a cloud? If they try, can any of them produce the economy of scale and community reach essential to success? Another is whether it’s simply too late for the EU to jump in; the cloud is too far along in both the US and China for anyone to catch up, even if they do it right.
In a sense, the problem the EU has with the cloud is a bit like the problem created by the “little ice age” in the 16th and 17th century. Established economic foundations, like a local agricultural economy and feudal system, were overturned by an external force—short-term climate change bringing much colder temperatures. While populations clung to established economic and political principals, they were unable to survive that way, and the result was a massive increase in global trade and the dawn of globalism, both economic and political. Today, we see established local computing and information storage, and the economic power of IT, threatened by the external force of US-dominated cloud computing. Is resistance to this futile, as it was to the cold? Is it possible for the EU to join the cloud revolution with sovereign cloud technology?
The Little Ice Age transformed the economy of Europe, and the cloud-and-AI age is doing the same, but we have to ask some new questions to get an answer to the questions of the last paragraph. The first question is whether Gaia-X or a successor would be effective. Another is what it would take to make it effective, and for it to stay that effective.
Does the EU want to pay the price, in social change, to even attempt this? Just creating and sustaining the EU in the face of nationalism is an issue, but replicating the wild-west nature of the US capital markets would require every country to cede financial regulation power to the central body. Then there’s the social questions. There’s a whole new tech elite class being created in the US by the cloud and AI. Europe would need the same thing, or any sovereignty initiative would flounder for lack of skilled personnel.
Can that class be created when the US market would serve as a magnet to any who developed skills? Will EU companies offer comparable pay, advancement, and startup opportunities? I don’t know, but I think Gaia-X so far suggests that these skills and opportunities aren’t there yet, and aren’t developing fast enough. The EU may agree that more time is needed; they just delayed their long-anticipated AI rules until 2027. If Gaia-X or a similar initiative is to work, a lot of earnest discussion is needed to get it on a track that leads to a positive outcome.
So what’s the answer? A race to be the first mover tends to generate surface technology advances without foundational elements that are really essential to building the future models of tech usage. In the cloud, in AI, the US has hyped the concept to the skies, but has failed to build the platforms needed to exploit the technology fully. The future of AI, and of the cloud, is way more than those hyped elements, and nobody in the US is racing to develop the extra pieces. Real-time, real-world, time-consuming technology of the sort I’ve blogged about in the past (and will blog more about shortly) is needed, and the US seems structurally incapable of developing it. Could the EU do that? I think it could, and should.
