Europe knows how to build strategic technology industries. We have demonstrated it before with the telecom industry.

European companies such as Ericsson and Nokia became global leaders in mobile infrastructure, and many smaller European companies built highly specialized products around the telecom ecosystem. Even today, large operators across the world, including in the United States, rely on European telecom infrastructure. In the 5G market, European vendors are not just competitive; they are among the most important choices available.

That success did not happen in an easy environment. Telecom in the 90s was shaped by national rules, state-controlled operators, complex regulation, standards bodies, procurement constraints, and political interests. European companies still built world-class technology. They pushed through fragmentation and regulation, created global products, and proved that Europe can compete at the highest level when it decides to build.

That is the mindset we need again for the next round of technology innovation.

AI has become a new strategic infrastructure layer, and it has already changed the way we plan and develop products and services, how we build and operate our processes, how we market them and how we use them. For Europe, the question is not whether we should use AI. We obviously will. The question is whether we want to be mainly consumers of foreign AI systems, or whether we want to help build a European AI industry that can stand on its own.

I believe European industry should deliberately spend money on European AI models and AI infrastructure, even when cheaper or “free” alternatives exist.

Not as symbolic patriotism but because strategic technology industries are built through real usage, real customers, real revenue, and real feedback.

“Free” AI Is Not Free

There are now, in May 2026, many capable Chinese AI models, including models such as DeepSeek, Qwen, Kimi, and others. Some of them are presented as open-weight or open-source. Some are offered for free or at very low cost.

From an engineering perspective, many of these models are impressive. Chinese AI researchers and engineers are clearly excellent. They are producing competitive models under difficult constraints, including restrictions on access to some Nvidia GPUs or GPU configurations and a domestic chip industry that is catching up quickly, with Huawei playing a leading role with its Ascend line of AI accelerators and AI processors.

If a powerful AI model is offered for free, somebody is paying for it. If the user is not paying, then the cost is carried somewhere else: through cheap credit, state-backed industrial policy, strategic subsidies, investor money, or national ambition. There is no such thing as free tokens forever or free intelligence at scale.

As much as most companies want to stay model independent, AI model usage creates dependency. Companies first use free tokens for experiments followed by building prototypes. Then they integrate the model into workflows, products, and processes. Eventually, the model becomes part of the architecture, the product, and the operating model of the company.

At that point, switching away is no longer easy. The tokens may have been free at the beginning, but they are not free at scale, and the dependency is real.

The Training Data Reflects National Interests

My concern is not that Chinese models are bad. My concern is that they are not neutral. They are trained and governed in a Chinese political, regulatory, and information environment. The available training data, including pre-training and post-training, the filtering, the reinforcement learning from human feedback, and the boundaries of what is encouraged or discouraged are shaped by the interests of the country and system in which these models are built.

The same is true, in a different way, for U.S. models. Large U.S. AI models are also shaped by U.S. commercial, legal, cultural, and national interests. They are usually more familiar to us in Europe because we have consumed American culture and technology for decades, but that does not make them neutral either.

We do not really know how the training data behind many models was chosen. We do not know which information was filtered out. We do not know which assumptions are embedded in the model. We do not know how the model will behave when its answers touch on areas of strategic importance. In consumer use cases, this may not always matter. In strategic industries like telecom, it does.

The Telecom Example: We Need AI Aligned With European Interests

I work in the European telecom software supply industry, specifically in the service assurance space. Our customers operate critical infrastructure. Our products help them understand network performance, detect anomalies, investigate root causes, and improve customer experience. These networks are the national communication backbones.

We use AI for automation, root cause analysis, anomaly detection, operational support, and eventually more advanced service assurance use cases. As a supplier, we need to be open to the models our customers choose. If a customer wants to use a Chinese model, we will support that. If a customer wants to use a U.S. model, we will support that. That is the reality of working with customers across different markets.

But when we use AI for our company processes and ultimately package AI capability into our own products, the decision becomes more strategic.

Which model should we rely on? Which model should we trust? Which model is aligned with European interests? Which model gives us enough transparency? Which model can we use without creating a dependency that may become a problem later? These are not abstract questions. We explored them concretely when evaluating the AI stack for GardenDesigner.ai — and the findings shaped both our technical and our strategic thinking.

At the moment, in May 2026, the European answer is still too weak. There are only a few serious European options. Mistral is the obvious name, but Europe cannot depend on one company alone. We need a much broader AI ecosystem.

In telecom, I would be very interested in a European AI model with deep telecom knowledge. A model that understands service assurance, alarms, KPIs, anomalies, topology, network operations, customer experience, and root cause analysis. Such a model would be highly valuable to companies like ours and to our customers. If a European vendor could offer that, I believe many companies would be willing to work with it, even if it was not perfect on day one.

European Companies Must Stop Waiting for Perfect Products

There is a contradiction in how many European companies think about technology sovereignty. We say we want European alternatives, but then we often refuse to buy from European suppliers until they are already as mature, as cheap, and as feature-complete as the global incumbents.

A company cannot become world-class without customers. AI model companies need usage. They need revenue. They need demanding enterprise customers. They need production workloads. They need feedback from real use cases.

If every European company says, “We support European AI, but we will use the free foreign model for now,” then the European AI industry will not mature. It will remain underfunded, underused, and strategically irrelevant.

This does not mean European customers should buy bad products. It does not mean AI companies should expect protection from competition. It means that in strategic areas, buying decisions have consequences beyond the next procurement cycle. If we want European AI companies to exist, we need to become customers early enough for them to grow.

AI Is No Longer Just for the Giants

One encouraging development is that AI is not only a game for the largest U.S. labs anymore. We see smaller companies, such as Arcee.ai in the U.S. and many Chinese labs, building competitive and useful models. We see companies with far less funding than the biggest players producing specialized models that are good enough for many real-world applications. We see Chinese companies producing impressive results with fewer resources than many people expected after the “DeepSeek moment.”

The opportunity is not limited to building the largest general-purpose chatbot in the world. There will be many valuable AI markets: small models, domain-specific models, private deployment models, coding models, telecom models, industrial models, compliance-aware models, and models designed for regulated environments.

Europe already knows the problems worth solving. Telecom, manufacturing, energy, healthcare, and industrial automation are not gaps in European knowledge — they are its strengths. That is exactly the kind of knowledge that turns a general-purpose AI model into something genuinely useful that can create value for consumers and industry.

We should not only ask whether Europe can build the next consumer chatbot. We should ask where Europe can build AI systems that solve serious industrial and operational problems better than anyone else.

Europe Needs a Full Strategic Technology Stack

AI is the most visible gap, but not the only one. Anyone working in European enterprise technology knows the feeling of scrolling through their stack and counting how many layers are controlled by companies outside Europe. The list is long.

As a European telecom supplier, we are willing to bet on European suppliers in strategic areas. We are interested in European-grown AI models, databases, code assistance tools, productivity platforms, and infrastructure components. In some areas, these suppliers exist. In others, they are still missing or not yet mature enough.

The new geopolitical situation should make this obvious. Europe cannot rely indefinitely on others to provide the most important layers of its digital infrastructure. We need partners around the world, but we also need our own capabilities.

Products, customers, revenue, and globally competitive companies create strategic autonomy.

Regulation Is Not a Reason Not to Build

Critics often use regulation as an excuse for lack of entrepreneurial ambition in Europe. They say the rules are too complex, the market is too fragmented, the AI Act is too limiting, or GDPR is too restrictive. Some of these concerns are real. But they can never be the reason not to build.

Entrepreneurs do not start with the reasons why something is impossible. They start with the will to build something new against all odds and through that they create true value.

Spotify is an example. The music industry was one of the most legally fragmented and commercially hostile markets imaginable. Spotify did not wait until a regulatory framework was established, but it built a new business model that has changed the music industry from the ground up. Daniel Ek, the founder of Spotify, negotiated, pushed, and built what became the platform for streaming music that connects artists seamlessly with their audience — and he did it from Stockholm.

Europe needs more entrepreneurs who are willing to push. Do not see existing regulation as a wall or an obstacle. Treat it as a target for innovation.

If a product creates real value for citizens, customers, and society, then build it. Improve it. Fight for it. Fight the good fight and let regulation adapt where it makes sense.

A Call to European Entrepreneurs and Customers

For European entrepreneurs, my message is simple: build.

Build products and services that create real value for consumers and entrepreneurs. Build things that help Europe become more productive, more secure, and more competitive. Build tools that allow Europe to finance its needs and protect its citizens. Do not wait until every rule is perfect. Do not wait until someone else takes the opportunity.

My own entrepreneurial journey started after two decades in the internet and telecom industry and eventually brought me to Polystar and Elisa Industriq — companies that chose to build rather than wait.

For European buyers, especially in strategic industries, my message is also simple: buy deliberately.

If we consistently choose free foreign models or strategic infrastructure because they are cheaper or more capable at that very moment, then we should not be surprised when European AI companies fail to scale. If we want European AI and strategic infrastructure, we need to create demand for it.

That does not mean ignoring quality and capabilities. It means understanding that quality and capabilities improve through real customers. It means being willing to work with European suppliers when they are credible, even if they are not yet perfect. It means seeing procurement as part of strategy, not only as cost optimization. And it means recognising that selecting the right vendor is only half the challenge — the harder problem is the execution discipline to integrate AI capability into real products and processes without losing momentum. That is a pattern we know well.

Europe has built strategic technology industries before. Telecom proves it. Let’s do it again together.