AI Gigafactories: Brussels' Computing Wonder or Just Hot Air?

KI & Automatisierung · 6. Februar 2026 · Ohiku Mose Guy

The EU aims to propel Europe to the top of the world with its own AI supercomputers. What this truly means for German SMEs – and where the catches are.

Last week, I was in a factory hall at an automotive supplier in Sauerland. Nothing special, at first glance. But then the production manager showed me an inconspicuous gray box on a multi-ton stamping press. This thing, he explained, spits out more data per second than an entire office processed per day in the 90s. Per second. On a single machine.

And this is precisely where the problem lies. We all talk about Industry 4.0, smart factories, and AI-driven processes. But we talk far too rarely about what all of this actually requires: computing power. And not just a little bit, not what your server in the basement can do. We're talking about absurd amounts of computing power. Power that we in Europe – let's be honest – largely buy from American hyperscalers like Amazon, Microsoft, and Google. Until now. Because in Brussels, they seem to have woken up and forged a plan that sounds so ambitious you have to look twice: they want to build their own "AI Gigafactories."

The Grand Plan: Europe's Digital Sovereignty on Steroids

So what exactly happened on February 6, 2026? The EU Council – at an almost breathtaking speed for Brussels standards – pushed through an amendment to the EuroHPC regulation. EuroHPC, that's the consortium that deals with Europe's supercomputers. Until now, it was primarily about scientific computing. Now, business is entering the game. And with full force. The regulation, which by the way is scheduled to come into force as early as January 20, 2027, dramatically expands the mandate: huge, specialized data centers – precisely these "Gigafactories" – are to be built across Europe, whose sole purpose is to train and operate the most advanced AI models.

The thing is: this isn't about a few more servers. It's about creating an infrastructure that can compete with the giants from the USA and Asia. An infrastructure on which a German mechanical engineer can optimize his digital twin, an Austrian logistics company can control its supply chains in real-time, and a Swiss precision manufacturer can elevate its quality control to a new level using computer vision. The idea is to build these centers through public-private partnerships and then make them accessible to industry. And – this is the crucial point – without our most sensitive production data having to leave the European continent. There's no getting around it: the issue of data sovereignty has finally reached the minds of politicians. A separate pillar for quantum technologies was also decided at the same time, but that's another story for another day.

The Bare Numbers: A Dangerous Imbalance

Why this sudden activism? A look at the numbers reveals it. In Europe, we have a massive two-tier society when it comes to AI adoption. While the big players are already diligently experimenting and implementing, SMEs – the actual backbone of our industry – are dramatically lagging behind. I see this every week when I visit companies. Some are talking about GPT-4o in control technology, while others are still struggling with the digitization of their delivery notes.

This table shows the dilemma (data based on the 2025 survey and my assessment). Almost 80% of SMEs are still on the sidelines when it comes to AI. Often not because they don't want to, but because the hurdles are too high: no staff, no idea where to start, and above all, the fear of monstrous costs for computing infrastructure and dependence on US providers. This is precisely where the Gigafactories are supposed to come in. A lever to lower the entry barrier. I doubt it will be that simple, but the approach is at least correct.

Politician's Prose and Reality on the Shop Floor

Of course, such a decision is accompanied by the usual political rhetoric. Nicodemos Damianou, some deputy minister from Cyprus, spoke of a "bold and swift step" that would strengthen "Europe's resilience, competitiveness, and sovereignty." That sounds good in a press release. Bold? Yes, perhaps. Swift? For EU standards, certainly. But what does that mean for you, the CEO of a 150-person company in the Black Forest?

They're putting a supercomputer there for us – fine and good. But who will train my people? Who will help me prepare my data so that the AI can even do anything with it? And what will the fun really cost me in the end when the initial funding runs out? These are the questions that keep me awake at night, not the number of teraflops in some data center in Finland.

— Dr. Anja Richter, fictional CEO of a medium-sized sensor manufacturer (based on my conversations)

And Ms. Richter hits the nail on the head. Providing infrastructure is only half the battle. It's a bit like building an Olympic pool for someone who can't swim. A great facility, but without a swimming instructor, without a lifebuoy, and without an explanation of how to stay afloat, that person will still drown. The real work – empowering SMEs – only begins after these factories are built.

The Global Context: An Expensive Race

This EU initiative must be seen in the right context. It is a desperate, albeit necessary, reaction to developments in the USA and China. While we in Europe debated data protection regulations (important, no question), others created facts. China alone installed almost 300,000 new industrial robots in 2024 – that's 54% of the world market! They are pushing automation with a ruthlessness that makes us dizzy here. And their AI companies are nurtured with massive state subsidies and a huge, protected domestic market.

On the other side of the Atlantic sit the true rulers of the AI world: NVIDIA, whose GPUs are in practically every data center. Google, Microsoft, Amazon, who control the cloud infrastructure. Companies like Tower Semiconductor and NVIDIA are teaming up to develop the next generation of AI hardware with silicon photonics for an absurd 1.6 trillion dollars. These are the spheres we are moving in. In comparison, a few EU Gigafactories – whose budget is not yet final – seem almost modest. But it is a start. A signal. Europe no longer wants to be just a paying customer in the Americans' computing power supermarket, but to open its own store.

But... isn't this putting the cart before the horse?

Let's be frank. I'm a journalist, not a cheerleader for EU initiatives. And with all due sympathy for the basic idea, I see massive pitfalls. First: the speed. The decision is there, the regulation will soon come into force. But anyone who has ever applied for a building permit for a garden shed knows how long it takes to plan, approve, build, and commission a "Gigafactory." Are we talking about 2028? 2030? In the AI world, that's an eternity. By then, NVIDIA will have three new chip generations on the market, and the models will be ten times more complex.

Second: access. There is talk of "flexible procurement rules" and "protection for start-ups." That sounds good. But in the end, the big players – Siemens, SAP, Airbus, the automotive groups – with their armies of developers and their huge amounts of data, will be the first at the trough. How will it be ensured that the SME from Buxtehude gets fair access to these resources? And not just to the computing time, but also to the know-how that is bundled there? I bet that will be the biggest challenge. And third, my favorite topic: energy consumption. An AI Gigafactory consumes electricity like a medium-sized small town. Where is that supposed to come from? And at what price? It's no coincidence that Google just concluded a massive deal for 100 MW of offshore wind energy in Europe, just to fuel its AI growth. Do we have this electricity? Sustainably and affordably? This question is often swept under the rug in Brussels.

What You Need to Do Now: A List for Doers

Okay, enough criticism. Waiting and complaining is not a strategy. The Gigafactories will come – in some form. The question is whether your company will be ready then. It's not about starting an AI department tomorrow. It's about setting the course today. Here's the homework:

1. **1. Conduct a data inventory – but radically honestly:** Forget big data. Start with 'right data.' Which three processes in your production are the most expensive, most error-prone, or slowest? And what data would you need to improve that? Focus on collecting exactly this data in good quality. Everything else is garbage.
2. **2. Define a lighthouse project (and budget it!):** Choose ONE problem you want to solve with AI. Not ten. ONE. Predictive maintenance on the most important machine, optical quality control for the most critical component, demand forecasting for the most expensive raw material. Allocate a fixed but manageable budget (e.g., €50,000) and a small, motivated team to it. Consider it paid further training.
3. **3. Forge partnerships that deliver:** Don't go to glossy consultants. Talk to the local Fraunhofer Institute, the robotics cluster in your region, or the technical university next door. That's where the application-oriented knowledge sits. Look for partners who have seen a machine from the inside.
4. **4. Appoint an 'AI caretaker':** You don't need a Chief AI Officer. But you need a person in your company who drives the topic. Someone who is curious, delves into the matter, and acts as a central contact person for external partners and internal teams. Give this person 20% of their time just for this topic. That's the best investment you can make.

My Conclusion: An Opportunity We Must Not Mess Up

Honestly: This initiative from Brussels is long overdue. It is the recognition of the brutal reality that technological sovereignty is the basis for economic prosperity in the 21st century. It's a gamble. An expensive gamble on the future of European industry. But it's better than idly watching us become a digital colony of other powers. The Gigafactories alone will not save us. They are just a tool – a very large, very expensive tool.

The real revolution must take place in people's minds. In the executive suites of SMEs. Away from German Angst, towards curiosity and the willingness to try things out. The greatest danger is not that AI projects will fail. The greatest danger is not starting any at all. I bet that in three years we won't be talking about how many petaflops the EU servers have, but about how many SMEs have actually managed to generate one more euro of margin with them. And that's all that matters at the end of the day. Let's get to it.