From market failure to a question of democracy: Rethinking digital power

Google, Amazon, Microsoft: three corporations that today control more infrastructure than most states can regulate. Anselm Küsters calls this behavioural power – and in his new book Small is beautiful 2.0, he has developed a counter-programme: open source, decentralised platforms, a theory of competition that takes democracy seriously as a protected good. Küsters completed his doctorate at our institute, conducts research at Humboldt University in Berlin and heads the Digitalisation Department at the Centre for European Policy (cep).

The central question of his book is: Can Europe take a third path – beyond American scaling logic and Chinese state control? A conversation about why E.F. Schumacher is more relevant than ever.

 

You advocate for policies that strengthen the small. What incentives do frugal AI, open infrastructures, and local digital innovation need in order to create real alternatives to the established platform giants?

The most important thing is that we stop equating digital innovation with size alone. We stare spellbound at the huge data centres and models of the US hyperscalers and think we just need to copy their approach. But Europe's strength lies elsewhere. Specifically, we need three things: First, radical promotion of open source as basic infrastructure. The International Criminal Court has just demonstrated how this could be used to break away from Microsoft's office software. Secondly, we need "real-world laboratories of decentralisation", i.e. money and legal freedom for local authorities and small and medium-sized enterprises to test local platform cooperatives or federated AI solutions such as Manufacturing-X. And thirdly, we need to overhaul the government's procurement policy: away from "Who is the cheapest (and has the biggest lobby)?" to "Who strengthens local sovereignty and open standards?". If the government acts as an anchor customer for OpenDesk or frugal AI models, a market will emerge that no start-up can create on its own.

That sounds like a fundamental power problem. You warn that large AI systems pre-structure decisions. Why does this combination of economic power and algorithmic control threaten citizens' sovereignty?

We are experiencing a shift in power that is difficult to grasp with our traditional concepts. In the past, a monopolist had pricing power. Today, a platform giant has behavioural power. When an algorithm decides what news I see, what product is recommended to me, or whether I am creditworthy, it deeply interferes with my autonomy. Finnish academic Vili Lehdonvirta calls this "cloud empires" because these corporations act like states. They set rules, administer justice, for example through content moderation, and levy taxes in the form of monopolistic fees. However, the problem is that these private entities are not accountable to us. If we take digital sovereignty seriously, we cannot simply privatise this power. We need "exception handling" – i.e. the ability to intervene in crises or in the event of wrong decisions – back in democratically legitimised hands.

During your doctorate at mpilhlt, you measured ordoliberal concepts in EU competition law using data analysis. What do you see as a result that traditional text reading overlooks?

It's a bit like the difference between a walk in the woods and a satellite image. Traditional reading shows the individual trees, such as the specific reasoning behind an ECJ ruling or the rhetoric of a European Commissioner for Competition. With text mining, by contrast, I suddenly see the whole forest and how it has changed over decades. In my work, I was able to show how the language of ordoliberalism, which is based on terms such as "competition for the merits", "freedom of competition" and "market structure", ran like a thread through European law – until these terms were almost abruptly replaced in the 2000s by a purely economic efficiency logic, the so-called "more economic approach". This becomes clearer when you look at thousands of documents as data points rather than just reading landmark decisions. Furthermore, when done methodologically robustly, such an approach somewhat dispels the notion that only the better argument prevails in individual cases – often it is discursive fashions and economic paradigm shifts that change the language of competition law just as much.

You are now working on your postdoctoral project at Humboldt University. When you comb through historical sources using digital methods, what do these patterns reveal about the recurrence of hopes and fears surrounding new technologies?

The debates surrounding the introduction of the steam engine, the telegraph, and the early internet are strikingly similar to our current discussions about AI. There is always this oscillation between utopian hopes of salvation and almost apocalyptic fears. Digital Humanities help to trace these discursive waves precisely – for example, through sentiment analysis, which can be used to search for emotions automatically. This is important for legal history, as it helps us understand that law never arises in a vacuum. If we recognise that we are currently in another phase of "technology panic," we can regulate the use of AI more calmly. This historical perspective is often missing in the hectic political debate in Berlin, Brussels, and elsewhere.

You call for a modernised ordoliberal "theory of harm". How should competition law address digital power in such a way that democratic risks become visible, and not just economic ones?

The classical ordoliberalism of the Freiburg School had an insight that is regaining relevance in times of libertarian tech billionaires and their dystopian visions of the future: private power is just as dangerous as state power. A private monopoly, for example over our data infrastructure, can threaten individual freedom just as much as a state dictatorship. In recent decades, some competition lawyers – especially in the US – have lost sight of this and focused solely on price. If the product became cheaper for consumers, efficiency gains were assumed and anti-competitive practices were accepted. In Europe, too, mergers and so-called "killer acquisitions" of tech start-ups were hardly prevented. But with Facebook or Google, the price we pay is "zero" only if we think purely in monetary terms. A modernised ordoliberal theory of harm would ask: Does this platform restrict my freedom of choice? Does it manipulate my opinion? Does it destroy the open market structure that we need for innovation and democracy in the long term? The EU's Digital Markets Act (DMA) goes precisely in this direction. It says: it's not about efficiency, it's about fairness and contestability. We must have the courage to recognise power for what it is, even when it is presented as a user-friendly app.

 

Anselm Küsters, Small is beautiful 2.0: Mit digitaler Dezentralisierung zu einer menschlicheren Wirtschaft, Herder Verlag, February 2026, 336 pages, EUR 26.00 (ISBN 978-3-451-03756-6)