On June 28, 2024, the French Competition Authority (“FCA”) released its opinion on the competitive functioning of the Generative artificial intelligence (otherwise known as “Generative AI” or “Gen AI”) sector. This opinion follows a self-initiated inquiry launched by the FCA in February 2024, which involved a public consultation that gathered insights from about forty actors. 

See here the very clear presentation made by the FCA during its press conference. 

The FCA’s opinion focuses on strategies by major digital players to consolidate their market power in the upstream AI value chain and leverage this power to grow in the sector. The FCA specifically examines practices by existing actors such as cloud infrastructure providers, and competition issues related to access to these infrastructures, computational power, data, and skilled labour.

Participants in the Generative AI value chain

The operators in the Generative AI value chain are:

• major digital companies: Alphabet and Microsoft are present across the entire value chain, while Amazon, Apple, Meta and Nvidia are present only at certain specific layers;
• model developers: for example, start-ups or AI-focused research labs, such as Anthropic, Hugging Face, Mistral AI and OpenAI. They have often formed partnerships with one or more digital giants, such as OpenAI with Microsoft and Anthropic with Amazon and Google. For the distribution of their models, they may adopt either a proprietary or open-source approach;
• IT component suppliers, such as Nvidia, develop graphics processing units (GPUs) and AI accelerators, which are essential components for training generative AI models;
• cloud service providers, including digital giants known as “hyperscalers”, such as Amazon Web Services (AWS), Google Cloud Platform (GCP) and Microsoft Azure, cloud providers such as OVHCloud, as well as specialist AI providers such as CoreWeave. The necessary computing resources may also be provided by public supercomputers (such as Jean Zay in France).

At the downstream level, many operators are marketing new services based on Generative AI to the general public (like ChatGPT), companies and public authorities and/or integrating Generative AI into their existing services (like Zoom).

Figure 1 Source: FCA

Competitive assessment of the Generative AI sector

The functioning of the Generative AI sector creates high entry barriers. 

A performing AI tool requires access to specialised AI chips, cloud services, large volumes of data, highly-skilled developers, and significant financing. These represent major barriers to entry.

Technical advances, like using synthetic data (those created by AIs), public supercomputers or open-source models may help reduce some of these barriers by partially replacing real data and easing the constraints of personal data use.

Major digital companies benefit from their activities in other digital markets.

Certain operators in markets linked to Generative AI may have privileged access to crucial resources for developing foundation models, including computational power, large volumes of data, and top talent. They secure computational power through bulk buying and preferential agreements with suppliers like Nvidia, and by developing custom AI accelerators like Google's TPUs. They also access vast data from their services and financial deals, such as Google’s $60 million annual agreement with Reddit (a US social news and discussion site). Furthermore, they attract talent with high salaries and innovation opportunities.

These companies also benefit fromvertical and conglomerate integration, ensuring access to users, businesses, and consumers. They leverage high fixed costs of training foundation models to achieve economies of scale and scope, using these models for diverse applications. Network effects enhance their models through user feedback, improving performance and creating new services. Companies like Microsoft integrate Gen AI into their product ecosystems, such as Copilot for Bing and Microsoft 365, and offer proprietary or third-party Gen AI models on their marketplaces to function within their ecosystems.

In other words, major digital companies already have a huge advantage over their competitors. 

In view of this market structure, the FCA identified a certain number of competition risks:

• risks of abuse by chip providers such as price fixing, production restrictions, unfair contractual conditions, and discriminatory behaviour. The graphics card sector, which was the target of an unannounced inspection in September 2023, is being closely scrutinised by the FCA’s Investigation Services;
• risk of lock-in by major cloud service providers; 
• risks related to data access (risk of refusal of access, discriminatory access, exclusive access, or on the contrary, use of content publishers’ data without their authorization); 
• risks related to access to a skilled workforce (wage fixing agreements, no-poach agreements, recruitment of entire teams…); 
• risks associated with vertical integration (limiting access to upstream resources, tying practices where major digital operators integrate Generative AI tools on their devices, self-preferencing…);
• risks associated with minority investments and partnerships by digital giants (these concerns are shared by competition authorities around the world, as evidenced by ongoing investigations into Alphabet, Amazon, Anthropic, Microsoft and OpenAI);
• risk of collusion using algorithms (see joint study of the FCA and the Bundeskartellamt[1]). 

Recommendations

The FCA concludes by issuing several recommendations to enhance competition in the Generative AI sector using existing regulatory frameworks in Europe and France. Key recommendations include:

• closely monitoring the development of cloud-based Generative AI services and considering designating such service providers as gatekeepers under the Digital Markets Act. In doing so, several problematic behaviours identified above would thus be prohibited ex ante;
• strengthening law enforcement in the sector (notably the SREN law in France which regulates competition in cloud markets);
• using competition law tools to address anti-competitive practices. The FCA encourages European authorities to remain vigilant and to act swiftly in the event of market abuses;
• encouraging innovation by improving access to computational power, in connection with the AI Act, set to come into force in 2026, establishing criteria for opening Generative AI models trained on public supercomputers;
• considering the economic value of data: agreements with rights holders should reflect the relative importance of the data according to the use case and specify in which circumstances the data may be used;
• increasing transparency regarding digital giants' minority, non-controlling investments in the sector or using the exact legal framework for concentrations (e.g. Article 22 of the Merger Regulation).

If you need more information or further guidance in this area, please contact Elsa Mandel and Mohamed Mamèche.

VISIT OUR COMPETITION & EU HOMEPAGE

[1] See here: https://www.autoritedelaconcurrence.fr/en/press-release/algorithms-and-competition-autorite-and-bundeskartellamt-publish-their-joint-study