DOWNLOAD FULL PDF

A small number of huge companies—Microsoft, OpenAI, and Alphabet—currently dominate in large language AI models. Because they have long had powerful positions, troves of data, and money to burn from existing markets, they could easily leverage themselves into the AI space. This has resulted in a worrying market-and-power concentration. Big market players are Google’s Gemini; Amazon; and Google-backed Anthropic’s Claude (OpenAI’s GPT-series that powers Microsoft’s Copilot), after multibillion-dollar investments in Open AI by Microsoft. OpenAI now runs primarily on Microsoft’s Azure.1 Greg Brockman, Ilya Sutskever, and Sam Altman, “OpenAI and Microsoft,” OpenAI, November 15, 2016, https://openai.com/index/openai-and-microsoft. And Mistral, the EU’s main attempt to compete with the American large language models, now also runs partly on Microsoft Azure after Microsoft’s €15 million investment in Mistral.2 Martin Coulter and Foo Yun Chee, “Microsoft’s Deal with Mistral AI Faces EU Scrutiny,” Reuters, February 27, 2024, https://www.reuters.com/technology/microsofts-deal-with-mistral-ai-faces-eu-scrutiny-2024-02-27. Both Microsoft “partnerships” are being investigated by the European Commission for being anticompetitive.3 Gian Volpicelli, “Microsoft’s AI Deal with France’s Mistral Faces EU Scrutiny,” Politico, February 27, 2024, https://www.politico.eu/article/european-commission-sets-its-sights-on-microsofts-ai-deal-with-frances-mistral.

These infrastructural agreements paired with large investments show the link between concentration in both the AI market and the infrastructures on which the AI systems run. Microsoft’s infrastructural power is leveraged into AI market share, as we see with the Microsoft-OpenAI partnership.

Concentration of Power in AI Is a Problem

What’s wrong with this scenario? First of all, if only a small number of large companies control the majority of the world’s AI systems and profit most, they gain the power to decide who gets access and under which conditions. This concentration in AI markets diminishes consumer choice, and erodes the autonomy for the people and organizations using AI, especially if they increasingly start relying on these systems. Ultimately, such market concentration has negative effects on accountability and it undermines our democracies if we rely on Big Tech’s shareholders to make decisions in the public interest. Second, amid worldwide geopolitical tensions, and with the US being a less stable political ally than in past decades, such dependence on a small number of US companies for AI and cloud services jeopardizes the EU’s strategic autonomy. For example, when Crowdstrike went down in July 2024, we witnessed how our dependence on a single company can plummet society into chaos: hospitals, trains, banks, and media outlets in the EU suddenly went dark.

Envisioning Solutions

We need to be ambitious. The EU must become digitally autonomous as soon as possible. In other words, the EU should no longer depend on large commercial tech companies for essential digital infrastructures, and the governance of these infrastructures should be kept as far as possible from commercial control. Beyond setting our own ethical standards for technology, as the EU has started doing with the AI Act, DMA, DSA, Data Act, and GDPR, the EU needs to build European alternative digital infrastructures. Access to an EU cloud could also benefit smaller ethical businesses and researchers running their AI.

One potential avenue for change that is currently not discussed much is the power of public procurement. With political commitment and investments through public procurement, we can change the current reality. Only then can we build the alternatives we want to see, instead of further reinforcing existing power structures.

Amsterdam has already committed to digital autonomy by 2030.4 “Agreement of the City Council with the Amended Initiative Proposal ‘Amsterdam Digitally Independent’ by Member IJmker,” March 16, 2024, https://amsterdam.raadsinformatie.nl/document/13943325/1/09012f978106234b. Today, the city already takes sustainability criteria and privacy into account in ICT public procurement. It also uses open-source software.5  Ibid. Before the end of 2024, Amsterdam will present different scenarios for a concrete procurement strategy, including “quick wins,” challenges, and necessary conditionalities for public procurement. 

We need similar ambition on an EU level. The Dutch government alone consistently pays Microsoft hundreds of millions of euros to run government services on Microsoft software. What if the Commission and all Member States came together and redirected the billions they already spend on Big Tech into secure, privacy-friendly, sustainable, and fast European digital infrastructure? We could invest billions of euros in ethical European initiatives rather than in Microsoft or Google.

Public procurement cannot be underestimated as a lever for change. Government contracts often represent a significant source of revenue for small and medium-sized enterprises (SMEs). Moreover, once a smaller business has a government stamp of approval and public investments have been secured, this attracts private investors and reduces the risk for private investments. 

The key here is strict criteria and conditionalities, so that public procurement actively shapes our European economy. The EU is lagging in this area: even the US Inflation Reduction Act channels increased investments to businesses that meet climate, clean energy, and social standards.

Taxpayers’ money should only go to the procurement option that is best for our society and economy—not only to the cheapest one. We therefore have to revise the European public procurement rules to ensure that sustainability, social standards, and privacy are binding criteria in tenders. Additionally, Member States should be able to consider their strategic autonomy and the long-term governance of the end product when purchasing, rather than being bound by the lowest price. Procurement procedures should be as simple and transparent as possible to enable small businesses and nonprofit initiatives to participate.

If we throw all our money at high-tech sectors such as AI without addressing the underlying infrastructural power dynamics, we ultimately reinforce our dependence on the same small number of large tech companies, doing very little for European SMEs.

For too long, we have allowed ourselves to believe that leaving digitization to the market would ensure freedom and prosperity for all. Now we are faced with the reality that the digital market is broken and led by a few companies who own the full digital stack, from basic essential infrastructure to consumer and government applications. 

As Europeans, we know how our digital infrastructure works, so we can fix it when it breaks. We need a Europe where we set bold goals so that we know which values lie at the heart of it. The ultimate goal of Big Tech is profits. Governments should lead the way forward and set an example. This starts with clear political commitment and binding criteria in public procurement.

If the EU wants to take the lead in certain AI sectors, we need to start building safe, sustainable, and ethical European digital infrastructure.

Before starting any race, you need to clearly define the starting gates and the finish line, and set your goals. Only then can you work toward achieving those goals with the right focus and training, and a safe, healthy path. This is exactly what the EU should do.