Details
The global race to build Artificial Intelligence infrastructure — chips, cloud computing, data centres — is driving an unprecedented concentration of corporate power. The rise of AI technologies is associated with an enormous, globally significant, unevenly distributed accumulation of power and capital: tech corporations such as Google, OpenAI, and Anthropic are competing to deliver “frontier AI models”, while Amazon Web Services, Microsoft Azure, and Google Cloud control two thirds of the growing global market in cloud computing services. The knowledge-intensive, capital-intensive production of powerful AI microchips is largely a matter of global strategic partnerships between just a few corporations: NVIDIA, TSMC, and ASML. Technically speaking, current AI models such as foundation models (FMs) and large language models (LLMs) — also known as generative AI — are based mainly on statistical optimization.
The directions taken by legacy approaches such as symbolic AI, or the formerly popular expert systems approach, considerably differ from the data-hungry computations of contemporary AI models, which incorporate neural networks patterned after the human brain. The rise of today’s models is linked with — and largely enabled by — a massive expansion in computing capacity and the exploitation of large data sets for commercial gain. The algorithmic, data-dependent mode of knowledge production that current AI technologies embody is compelling adaptation across a growing assortment of economic domains and producing new, complex forms of uncertainty — whether in relation to jobs, wages, or border controls. Some have even expressed the worry that AI could trigger a stock market crash.
A Global Infrastructure of Dependency
A critique of AI focused exclusively on the model and its applications falls short, however. A look at the value and supply chains needed to produce AI systems makes it clear that contemporary AI is not made up of chatbots and other software alone. And that it will hardly be possibly to curb AI’s impact on systems of dominance simply by calling for increased “fairness” and more stringent data protection.
AI supply chains are made up of hundreds of firms and thousands of distinct processes and products, with the latter produced and shipped via an opaque, extremely unequally distributed economic infrastructure: a global web of production facilities, electronics, data, energy, human labour power, and natural resources. These chains can be understood as contested forms of the social division of labour, through which claims to resources, labour, and global hegemony are negotiated — whether via disputes over the construction of and operation of energy-intensive data centres, struggles for better working conditions, or conflicts surrounding the environmentally destructive extraction of critical minerals.
Chips, Clouds, and the New Geopolitics
New geopolitical tensions — particularly as China rises and an economic war between China and the US begins to unfold — have given prominence to digital infrastructure and “frontier technologies” and led to new competitive dynamics at the global level. In the race to achieve technological dominance, governments are increasingly focused on creating attractive investment conditions and economic opportunities for AI firms while also building up their own technological capacities.
If AI is understood as a general-purpose technology, it possesses strategic relevance for both state actors and firms and is key to shoring up geopolitical dominance.
If AI is understood as a general-purpose technology, it possesses strategic relevance for both state actors and firms and is key to shoring up geopolitical dominance. In contrast to “classical” geopolitical conflict, which revolved around territory, the primary focus today is on achieving the power to control and shape global networks and infrastructure. At the nodal points of this global infrastructure — for example semiconductor technologies, AI chips, and access to cloud computing resources — relationships of mutual dependence can be leveraged in the pursuit of the national interest. New strategic partnerships are continually being formed with the goal of promoting the growth of AI technologies and improved supply chains, not least because the high volume of investment in the AI industry has made it critically relevant to global markets.
The Tech Oligarchy Enters Politics
In all this, the global AI ecosystem remains under the vertical dominance of the proprietary markets, value-creation strategies, and “black box” architectures of US and Chinese technology companies (Big Tech). That does not mean that other tech companies and suppliers are unable to successfully implement strategies for value creation, for example with respect to the development of AI models of their own. However, in addition to their position as providers of digital services, AI systems increasingly allow Big Tech to capture the knowledge needed to plan and innovate, leading to enormous power imbalances and relationships of economic dependency not only between companies, but between entire regions, particularly in the Global South.
AI innovation has increasingly been financialized, with the result that alongside tech firms, individual investors controlling large amounts of capital can exercise an unprecedented degree of power over the ways that AI infrastructure is shaped.
Moreover, AI innovation has increasingly been financialized, with the result that alongside tech firms, individual investors controlling large amounts of capital can exercise an unprecedented degree of power over the ways that AI infrastructure is shaped. Especially in the case of the US, these developments are usually discussed with reference to the tech oligarchy: tech company founders and CEOs such as Elon Musk, Peter Thiel, and Mark Zuckerberg who use their economic and infrastructural power to influence political institutions and decisions in the furtherance of their own economic aims, employing lobbying tactics, lawsuits, legislative initiatives, electoral campaign financing, and direct cooperation with political elites.
While often prominently endorsed when the declared object is “digital sovereignty” or “strategic autonomy”, the increased allocation of resources to AI-relevant infrastructure in regions and countries that trail behind in this intensified competitive regime does not necessarily lead to a complete technological decoupling but, on the contrary, to a deeper integration into either the US or the Chinese sphere of technological influence.
To take a recent example: the AI Impact Summit in New Delhi in February 2026 (which was not attended by any Chinese delegation) saw India join the US-led Pax Silica initiative. The Pax Silica initiative is primarily focused on securing strategically relevant critical minerals and semiconductor supply chains. In the particular case of India, this is accompanied by massive involvement of companies such as NVIDIA and Microsoft in the expansion of AI infrastructure, but also by new partnerships such as those between the Indian giants Tata and Infosys and OpenAI and Anthropic.
What we are seeing here are deliberate steps toward creating new markets and cementing relevant AI infrastructure by expanding it — a tendency that can also be observed in Europe.
A Values-Based AI Policy for Europe?
The AI policy of the European Union continues to be described as a “values based” alternative to conditions in the US and China. In this narrative, the EU is said to have particular power in setting rules that have global impact. The hope here is that in the future AI applications will be developed in Europe in accordance with certain minimum ethical standards. Despite the limited scope for EU institutions to coordinate industrial policy — which is developed and implemented primarily at the member state level — the ambition, frequently invoked in European AI policy circles, of making digital technologies more “human-centric”, “trustworthy”, and “sustainable” should not be completely discounted. It has in fact led to political compromise solutions that are now coming under renewed fire.
Dependence on US cloud services are now viewed as a potential risk, given the Trump administration’s demonstrated willingness to force open new markets by exploiting relationships of economic and technological dependence.
One example of such fragile compromises is the AI Act, which came into force in August 2024. It was presented as a milestone in AI regulation and hailed as the first comprehensive regulatory framework of its kind. The framework divides AI models into risk categories according to their potential to cause harm, a move that has consequences for how they can be used and for issues of liability. The act can be viewed as a continuation of efforts to set legal limits to the market power of digital technology companies.
At the same time, the EU’s Digital Markets Act and its Directive on Platform Work are testament to the fact that these efforts are also aimed at ensuring better working conditions and consumer and data protection. Meanwhile, public debate over cybersecurity and the vulnerability of digital infrastructure has also been gaining momentum in the EU, at the latest since the Russian invasion of Ukraine four years ago. Accordingly, priorities articulated through political programmes have also shifted. Dependence on US cloud services are now viewed as a potential risk, given the Trump administration’s demonstrated willingness to force open new markets by exploiting relationships of economic and technological dependence.
The EU’s Deregulation Agenda
However, in 2025 the European Parliament also took a step to the right, enabling an authoritarian expansion of Europe-wide use of AI surveillance powers, something that bodes ill for the future. Under pressure from tech firms and the US government, but also with encouragement from lobby groups such as Business Europe, explicit calls to dismantle “impediments to competition” and “bureaucracy” are being heard yet again.
This deregulation agenda — taking up well-worn neoliberal principles, and almost entirely based on the false opposition “regulation versus innovation” — is in evidence in what is known as the “Digital Omnibus”. This legislative package was introduced in late 2025 to provide “technical streamlining” to the EU legislation on digital technologies. Part of it consists in legislation aimed at watering down the AI legislation that had only just been passed, in particular as regards the handling of high-risk AI systems, as well as the adjusted rules on data use, data protection, and cybersecurity.
The EU's deregulation agenda consists in legislation aimed at watering down the AI legislation that had only just been passed, in particular as regards the handling of high-risk AI systems, as well as the adjusted rules on data use, data protection, and cybersecurity.
These changes were criticized by civil society organizations as severely undermining social rights and environmental protections. In her role as president of the European Commission, Ursula von der Leyen, by contrast, summarized the attack as a beneficial “simplification” that perfectly matches the EU’s new “AI first” line: parallel to the deregulation, Europe, like other jurisdictions, is to see more and more state resources mobilized with the aim of furthering a new politics of AI development.
Meanwhile, business associations representing companies involved in AI development never tire of claiming that international comparisons show European infrastructure spending to be too low. According to consulting firm McKinsey, private-sector AI investment in the US over the years 2020–2024 was about five times as high as private-sector AI investment in Europe. But in Europe as elsewhere, private capital plays an ever larger role, now representing a key component in a market-based strategy for the financing of “European AI”.
InvestAI and the European Silicon Valley
This is exemplified by coalitions currently being promoted by the European Commission, including the InvestAI initiative, touted as the world’s biggest public-private partnership for the development of “trustworthy” AI. The initiative brings together member states, the European Investment Bank (EIB), and private-sector organizations associated with the EU AI Champions Initiative (over 70 companies, including Airbus, Mercedes, Siemens, and SAP) with whose help €200 billion are to be raised for the development of what have been dubbed AI gigafactories, where AI models are to be trained. This gives private investors significant leverage in a future AI ecosystem that is expected to crystallize as a European Silicon Valley.
Given that Europe, according to the title and mission statement of the European Commission’s current action plan, the “AI continent”, still has no large cloud computing provider of its own, small and medium-sized European businesses and, above all, start-ups are seen as key agents of the AI transformation. The declared goal is for AI to be widely deployed and used, with companies being urged to adopt an “AI-first mindset”. “Computing power, data, and talent” are to be brought together in newly emerging AI factories. According to Ursula von der Leyen, it is imperative to adapt as quickly as possible to the systemic demands of “AI innovation” and “AI acceleration”, to ensure that “the future of AI is shaped in Europe”.
Technological Dependency
Paradoxically, it would appear that as part of the restructuring currently being launched, old relationships of technological dependence are to be swapped for new ones, but old ones deepened as well: instead of rigorously opposing the power of US tech corporations, interim technical solutions are being sought. The main actors who stand to benefit from the latter, however, are just those US corporations, with slightly adjusted product portfolios and business models. The new geopolitically fuelled focus on the expansion and provision of AI infrastructure gives them an opportunity to make optimal use both of the head start that their data-sets and integrated software and hardware architectures have allowed them to gain in the cloud computing market, and of their lobbying resources.
Very much in line with the time-honoured strategy pursued by digital platforms of offering key platform services as digital, modular solutions and thereby tying customers to the provider’s ecosystem (“lock-in”), the new (old) slogan is “sovereignty as a service”.
The irrationality of these technological conditions has to be called out. That includes remembering that neither increasing the amount of “AI innovation” across the board, nor “accelerating” it can stand in for social progress.
It is primarily Microsoft, Amazon, and Alphabet (Google) who are taking strategic positions in the context of the debate in Europe over “digital sovereignty”, reinterpreting the topic according to the logic of the Californian Ideology, and reducing it to technical, legal, and infrastructural aspects. Instead of enabling the actual exercise of sovereignty, the platforms provide “sovereignty” under conditions that they set, a service that can be acquired, configured, and optimized.
Thus, rather than terminating technology companies’ network centrality, privatized provision and acquisition of digital infrastructure promotes tech companies’ power. This winner-takes-most dynamic incurs high costs, given that most deployment and utilization of AI in Europe involves knowledge and value flowing to Big Tech corporations. Hence current European efforts in no way lead to an automatic end to relationships of technological dependence, not to mention to sustainable ways of tackling the ecological crisis or better working conditions. These perspectives appear, if at all, only as afterthoughts to the present agenda.
The irrationality of these technological conditions has to be called out. That includes remembering that neither increasing the amount of “AI innovation” across the board, nor “accelerating” it can stand in for social progress.
On the contrary, scepticism is warranted regarding both the black boxes “artificial intelligence” or “innovation,” and the competition-oriented benchmarks that are often appealed to (such as the Global AI Index). Rankings of that kind distract from the fundamental question: to what degree should integration into the package deal of the global AI ecosystem, including all of its negative dominance effects, be renegotiated? Instead, the debate is about how to best to fit ourselves into systems of private, capitalist appropriation.
This article was first published in German in the journal LuXemburg
Translated by Marc Hiatt and Sam Langer for Gegensatz Translation Collective.
