From left to right: Konrad Ost (Bundeskartellamt), Blanche Savary de Beauregard (Mistral), Vincent Champain (Framatome), Clark Parsons (IE.F), Elodie Vandenhende (Autorité de la concurrence); Picture Source: Agata Hidalgo
We are currently experiencing a Cambrian explosion in AI, reminiscent of the dot-com boom of the late 1990s, but at a far greater intensity. The pace of AI advancements feels exponential, with each product iteration rapidly pushing boundaries and creating new opportunities. Similar to the internet, AI's potential extends to nearly every sector, making disruption or transformation almost inevitable. This is a great chance for European tech.
The Infrastructure of AI: Major Players and their Advantages
Unfortunately, established big tech companies are leveraging their already existing market power. Because of this, European startups wanting to use and scale AI struggle when it comes to accessing high-quality data, infrastructure and talent. To level the playing field, a variety of measures needs to be employed, also to accommodate for the different markets that play into the training of AI. The biggest barriers for competition between European innovators and Big Tech are multifaceted:
Data Dominance: Access to vast amounts of quality data is crucial for training and building LLMs. Tech giants such as Google have a significant lead here, as they leverage already existing data pools and strike exclusive deals and partnerships, which amplify their competitive edge.
Cloud Advantage: The companies behind dominant cloud providers (Azure, AWS, Google Cloud) are consolidating their power through vertical integration (combining access to users, data, cloud, data centers, researchers etc) which helps them scale more effectively, making it difficult for new entrants to break through. In addition to that, some dominant cloud providers hold power over their users by, for example, charging fees over migration of data.
Investments, Partnerships and Killer Acquisitions: Big tech companies have taken to venture capital to strengthen their market position with Amazon, Microsoft and Google making up two thirds of VC in gen AI in the US. Like this, startups depend on big tech early on which strengthens their hold on the market and prevents competitors from emerging.
Talent Shortage: AI expertise is scarce and highly valued. Companies like Google and OpenAI can command immense investments, with talent acquisition deals in the tens of millions. This includes “acqui-hires” where startups are acquired mainly to secure their human capital.
Capital: Building foundational models requires substantial financial resources. The training costs are enormous, and the half-life of models becomes shorter and shorter.
Hardware Dependency: Training advanced AI models requires immense computational power, benefiting chip manufacturers like NVIDIA, which have become highly valuable and powerful as a result. Europe is lagging behind when it comes to data centers and chip production and is dependent on providers from third countries.
Europe’s Chance in AI
While the power of these dominant players seems overwhelming, not all hope is lost. What Europe needs is the chance for more small and new entrants to be able to find their corner in the market. In order to give them a chance, the continent has to continue to build its own infrastructure. Until then, many European startups are already leveraging the AI and cloud services that big US players provide. They build innovative solutions on top of this foundation.
Using the DMA to boost competition
The problem becomes apparent when looking at the terms and conditions of these services. Companies such as Google and Amazon offer free migration from their clouds to others while Microsoft is still reluctant to do so. In fact, the US Federal Trade Commission is planning to investigate the company’s practices, especially Microsoft 365’s incompatibility with other cloud products and the fees it raises on those who leave the service. For proper competition in AI, contracts that do not feature exclusionary clauses and grant fair access to cloud services are essential. Lock-in effects, exorbitant pricing rules and other predatory conditions stifle competition and thus innovation.
Fortunately, the EU has the advantage of the Digital Markets Act which it should use as a strong tool to boost competition. The first step would need to be to designate dominant big tech companies as gatekeepers under the Core Platform Service of Cloud Computing Services under Article 2.2 of the DMA. Following Article 6.6 and 6.7, this would ensure that they have to give companies the chance to move their data to other services if needed, thus preventing lock-in effects.
Preventing self-reinforcing market dominance through access to data
Even if such lock-in effects are avoided, the initial access to data is one of the most important aspects for AI training. The example of Google shows how a monopoly in one area (Search) can lead to a major advantage in another area (AI). Some even argue that the company is using the same methods in the field of AI as it was doing in search, such as making exclusive deals with companies to receive their training data and extending their dominant position even more. This is why it is important to level the playing field by granting access to big tech’s high quality data in order to foster competition. In the case of Google, this could be done by extending the non-compliance investigation under the DMA that is already being carried out against the company on the topic of self-preferencing in search.
Easing EU reporting burdens
European companies are not only locked-in by the power of these giants, they are also held back by the burden of EU reporting requirements. Regulations like the GDPR and the new AI Act are placing a high compliance burden on companies that are just starting out.
The burden that the EU places on businesses through these regulations also skews competition in favour of big companies who have the means to employ legal teams and might even risk fines. Smaller companies and startups are unfairly disadvantaged having to spend precious resources on compliance rather than research and development. Reevaluating redundant reporting has been deemed central to the new EU Commission and from a competition and competitiveness perspective, these regulations should be the first ones to be evaluated.
How to move forward
AI is not just a technology topic but has become a matter of national security, energy consumption, and environmental impact. The current use cases of generative AI within companies are mostly in the information industry. The next step would be to unlock the real added value for fields like manufacturing, compliance and others which should materialize at the intersection of domain-specific knowledge and new language models. Companies being able to use models based on their own domain knowledge on their own premises will also benefit European tech sovereignty.
Leveling the playing field is the first step to fully realizing this potential. This is why Europe should use the measures it has created for its industries to stay competitive. Ensuring fair competition requires action, especially to curb predatory practices, provide access to data and foster a more level playing field in AI. In addition, the EU needs to reevaluate the burden that regulation places on smaller businesses and startups.