DOWNLOAD FULL PDF

The European Commission’s proposal for boosting AI combines initiatives to promote its supply and demand.1 European Commission, “Communication on Boosting Startups and Innovation in Trustworthy Artificial Intelligence,” Policy and Legislation, January 24, 2024, https://digital-strategy.ec.europa.eu/en/library/communication-boosting-startups-and-innovation-trustworthy-artificial-intelligence. “AI factories” with public supercomputers and services for training models and developing applications are expected to galvanize supply. And, for stimulating AI appetite, the EC proposes “GenAI4EU,” a grouping of scattered grant schemes to expand adoption in an array of industries, science, and the public sector.

A straightforward critique would put eyes on the recommended investment: a total of €3.3 billion for the “AI Factories” and half a billion for “GenAI4EU” by 2027. Although these figures may seem astronomical, in early October 2024 Microsoft announced that it will invest $2.7 billion in Brazil alone between 2024 and 2027 for cloud and AI infrastructure, and $4.7 billion in Italy over the next two years.2“Microsoft Invests €4.3B to Boost AI Infrastructure and Cloud Capacity in Italy,” press release, Microsoft, October 2, 2024, https://news.microsoft.com/pt-br/microsoft-announces-14-7-billion-reais-investment-over-three-years-in-cloud-and-ai-infrastructure-and-provide-ai-training-at-scale-to-upskill-5-million-people-in-brazil. And in 2023, two-thirds of the $27 billion invested in generative AI startups came from this giant, Amazon, and Google.3George Hammond, “Big Tech Outspends Venture Capital Firms in AI Investment Frenzy,” Financial Times, December 29, 2023, https://www.ft.com/content/c6b47d24-b435-4f41-b197-2d826cce9532.

A relatively meager investment is, however, the least of the EC’s problems. While its proposal aims to create a competitive EU market for AI, it ignores that the AI global value chain is not a market but a planned sphere. Unlike old monopolies, Big Tech controls beyond ownership, using its concentrated data, AI talent, and digital infrastructure to plan the whole chain, dictating how and what AI models and applications are developed.4Cecilia Rikap, “Dynamics of Corporate Governance Beyond Ownership in AI,” Common Wealth, May 15, 2024, https://www.common-wealth.org/publications/dynamics-of-corporate-governance-beyond-ownership-in-ai. They do so acting as corporate venture capitalists, dominating the global AI knowledge and innovation network and ruling from their clouds.5 Cecilia Rikap, “Varieties of Corporate Innovation Systems and Their Interplay with Global and National Systems: Amazon, Facebook, Google and Microsoft’s Strategies to Produce and Appropriate Artificial Intelligence,” Review of International Political Economy (June 2024): 1–29, https://doi.org/10.1080/09692290.2024.2365757. These are not just “factories,” but planned spheres with rules clearly set by Big Tech. As thousands of organizations coproduce AI in these spaces, the three giants profit disproportionately.

Recipes that Reinforce the Status Quo

The EC’s industrial policy and pro-competition playbooks will backfire. As odd as it may seem, cloud giants embrace competition and innovation in the AI value chain. AI’s pattern of innovation is characterized by an entrenched and established core of Big Tech surrounded by a turbulent periphery.6 Cecilia Rikap, “Intellectual Monopolies as a New Pattern of Innovation and Technological Regime,” Industrial and Corporate Change 33, no. 5 (October 2024): 1037–1062, https://doi.org/10.1093/icc/dtad077. A flourishing startup community expands the periphery of companies producing parts of the chain. And as competition—turbulence—increases, whether for AI modeling or in any AI application field, the better it will be for Big Tech. Not even a promising startup like OpenAI will dream of debunking the giants, since it will have to focus on systematically winning the innovation race in its gear of the chain. Meanwhile, only Big Tech produces AI autonomously, retaining a panopticon view and value-chain bottlenecks.

While European startups may train their models on EU supercomputers, they will in any event be pushed to sell where demand goes, thus on Big Tech clouds. On top of offering infrastructure as a service, their clouds are supermarkets of computing services that spare small companies from having to develop every single line of code for each solution. Even Meta, which trains its models in-house, offers them as a service on Amazon, Microsoft, and Google clouds.

GenAI4EU is made of Horizon Europe calls and public-private partnerships. It is highly unlikely that these instruments will divert demand from the cloud. Big Tech has armies of outsourced companies and has developed dozens of strategic partnerships. Google has a list of eighty strategic partners in Europe, mostly leading corporations like LVMH7“LVMH and Google Cloud Create Strategic Partnership for AI and Cloud-Based Innovation,” Google Cloud, press release, PR Newswire, June 16, 2021, https://www.prnewswire.com/news-releases/lvmh-and-google-cloud-create-strategic-partnership-for-ai-and-cloud-based-innovation-301313307.html. and Renault,8 “Renault Group and Google Accelerate Partnership to Develop the Vehicle of Tomorrow and Strengthen Renault Group’s Digital Transformation,” Renault, press release, November 8, 2022, https://media.renaultgroup.com/renault-group-and-google-accelerate-partnership-to-develop-the-vehicle-of-tomorrow-and-strengthen-renault-groups-digital-transformation. which then hire startups for specific solutions. I would pay to see all these European companies breaking existing agreements for a Horizon grant.

Another area of concern picked up by the EC proposal is Europe’s scant public and private funding for AI startups and scale-ups, which either fail, or—like the French Mistral AI—end up partially funded by Big Tech.9 Rikap, “Dynamics of Corporate Governance Beyond Ownership in AI.” But things are even more complicated because European corporations support US startups. By February 2024, 59 percent of the companies receiving SAP’s venture capital were based in the US and only 11 percent were German. It is unclear how investors will prioritize European companies. Anyway, without breaking up their relationship with Big Tech, pouring more public or private money into European startups will—for the reasons I have explained above—end up favoring the cloud hegemons.

Both in terms of digital infrastructure and investment, the EC does a good job identifying Big Tech as the problem, but fails to fully understand the complexity of Big Tech’s AI stranglehold. Something similar happens with talent, which is seen by the EC as crucial for developing independent AI. However, the EC’s proposal to expand talent is shortsighted. It suggests more collaboration with European startups and universities, while neglecting that the most talented AI scientists and engineers work either for Big Tech or for their satellite startups. These people will not massively leave their jobs to apply for an ERC or Marie Skłodowska-Curie grant. What is worse, the talent that has remained in academia, in Europe and elsewhere, either works part-time for Big Tech or is funded by Big Tech.10 Rikap, “Varieties of Corporate Innovation Systems and Their Interplay with Global and National Systems.” Working closer with such academics means working with (for) Amazon, Microsoft, and Google.

In short and inadvertently, the EC proposal is a one-way ticket to play Big Tech’s game, favoring the companies that it accurately aims to keep at bay.

Reimagining Sovereignty in the AI Age

Considering the scale and centrality of Big Tech’s monopolized positions in data, talent, and the cloud, some might call for a European champion. Besides being unrealistic, the problem is not these giants’ nationality but the fact that Big Tech companies are intellectual monopolies;11 Cecilia Rikap, Capitalism, Power and Innovation: Intellectual Monopoly Capitalism Uncovered (Abingdon, UK: Routledge, 2021), https://www.routledge.com/Capitalism-Power-and-Innovation-Intellectual-Monopoly-Capitalism-Uncovered/Rikap/p/book/9780367750299. they appropriate value by capturing knowledge and data coproduced with or simply produced by many others. Neither can solutions spring from promoting competitive markets in the void. Parts of the AI value chain require scale and are too sensitive to be left in private hands. Private accumulation is companies’ DNA; thus nothing assures that we will get the AI that society and the planet need if AI is left in private hands.

Instead of—unsuccessfully—chasing US and Chinese Big Tech by creating a European AI market or champion, the EC should show that another AI is possible and desirable. Given AI’s centrality, associated risks, and tendency toward monopolization, the only way to do this is by democratically planning its development and use. But this is the main weakness of the EC proposal: its complete lack of desire to plan—to steer, coordinate, and shape—what AI is produced, why it is produced, and who it is produced by.

An EU-led AI plan should comprise a truly public cloud that is not only a factory but also a space where AI and other computing solutions can be accessed. And, along the lines of Brazil’s AI plan,12 “Brasil Launches a USD 4 billion Plan for AI and Prepares Global Action,” G20 Brasil 2024, July 30, 2024, https://www.g20.org/en/news/brasil-launches-a-usd-4-billion-plan-for-ai-and-prepares-global-action. it should also include foundational AI models governed as a commons and a strategy to bring talent back. The plan and its associated public technology—the cloud and foundational models—should remain under the oversight of a new European institution independent of corporations and individual governments. This institution should safeguard human and civil rights and assess AI development and use against energy and water consumption. It should also take the bold step of recognizing that an AI plan that puts the people, the planet, and democracy first requires that economic gains and efficiencies take a back seat. Only then will such a plan be a steward of the sovereignty of states and peoples in the digital age.