The transformative potential of generative AI extends across various industries, reshaping the way we live and work. These changes will unfold rapidly, exerting a significant impact on competition and consumers. Consequently, there is a pressing need to proactively monitor and regulate these changes from the outset, employing the most effective regulatory frameworks.
The development of many generative AI models relies on a specific set of key inputs, commonly referred to as AI building blocks. As such, the initial competition-related considerations revolve around these inputs, necessitating a comprehensive understanding and oversight to ensure fair and effective market competition.
AI Building Blocks
The essential building blocks of AI systems are data, talent, and computational resources. Control over one or more of the AI building blocks can significantly impact competition within generative AI markets. Beyond the AI field, competition regulators have already developed theories on foreclosure, self-preferencing and discriminatory behaviour concerning the control of inputs.
A crucial input for the development of generative AI models is data. Specifically, the quality and volume of data are crucial factors in the successful learning process of the AI model. Hence, more established companies, especially the ones that also own digital platforms may benefit from their access to user data accumulated over the years. Even when lawfully collected, a company’s control over data may serve as a barrier to entry, potentially limiting competition.
To compete in the AI space, companies must possess professional expertise and engineering experience i.e., talent. The development of generative AI models requires a deep understanding of machine learning and natural language processing—a relatively rare skill set. Consequently, companies may be inclined to lock in skilled workers through non-compete clauses, hindering their mobility. Furthermore, the companies could collude with competitors, for example, agreeing not to hire each other’s workers at all, or not to offer them a higher salary to change the job. Such actions could adversely impact competition by preventing new companies from entering the AI market, where success relies on the ability to hire highly skilled workers.
Computational resources constitute the third essential input in the AI generative market. These resources are crucial for processing data, training the model, and deploying the generative AI system. Computational resources can take the form of dedicated hardware, such as specialized chips, which are costly to obtain and maintain. Alternatively, companies can utilize cloud computing services which provide computational resources on demand. However, cloud computing services are expensive and offered by a limited number of firms that have already restricted their availability due to high demand. As firms in highly concentrated markets are more prone to activities of unfair competition, the cloud computing market and multiple processors market could also be under the scrutiny of competition watchdogs. For example, cloud computer providers could exploit the generative AI companies’ needs for computational resources by charging high egress fees.
Possible Competition Concerns
Companies that currently control crucial inputs for the AI generative market or adjacent markets such as the production of specialized microchips and the cloud computing market, may be inclined to leverage their existing power to gain control over the emerging AI generative market. They might employ strategies such as linking new generative AI products with their existing core products (tying) or offering bundled packages of multiple products (bundling), potentially distorting competition. Incumbents could also engage in self-preferencing activities, exclusive dealing, or other discriminatory practices against new entrants, further stifling competition.
First movers in this space could benefit from network effects, allowing them to rise to a dominant position or to concentrate market power. As generative AI models improve with increased user interaction, first movers can benefit from a feedback loop that enhances the performance of their models. This may result in a concentrated market with limited opportunities for new entrants to compete.
M&A activity could also serve as a tool to concentrate market power. Larger players may acquire critical applications, cutting off in this way competitors from core products, or directly acquiring rivals entering the market with the intention of shutting down competing products (so-called killer acquisitions).
The information in this document does not constitute legal advice on any particular matter and is provided for general informational purposes only.
