A Competition Policy for Cloud and AI

As European leaders focus on boosting Europe’s competitiveness, they must ensure businesses use technologies like artificial intelligence (AI) to boost competition across the economy. AI can do so by disrupting incumbents in the tech sector and by helping firms in many sectors become more productive.

To maximise the economic benefits of AI foundation models, competition authorities must ensure that competition thrives throughout the AI value chain – so AI remains as cheap, high-quality, innovative and widely available as possible.

This issue paper provides an overview of how effective competition between providers of AI foundation models is functioning today. It focuses on upstream inputs, in particular the provision of computing power and data for AI foundation models, rather than downstream uses of AI – such as competition between applications and devices that deploy AI, or future markets that might develop using AI, such as markets for AI agents.

Authors Zach Meyers and Marc Bourreau observe that:

• The AI sector appears to be competitive in terms of both static metrics of market power and dynamic competition analysis.
• The provision of accelerated compute should be analysed separately to existing markets in the general cloud computing sector, because it is subject to unique competitive dynamics.
• Some chokepoints, like access to data, may require targeted regulatory interventions to protect competition between AI developers.
• The concentrated nature of some parts of the AI value chain, and the strong position of the hyperscalers across multiple parts of the value chain, also create the possibility of future harm to competition, which requires close scrutiny. In particular, authorities should avoid situations where AI firms or customers are ‘locked in’ to particular suppliers.
• However, many firms in the sector are adopting relatively open approaches, which give firms and customers choices. It is not yet clear whether (or when) they will have the incentive or the ability to adopt more closed business models.

They recommend that policy-makers examine carefully how future market changes (in particular any move away from openness in the AI sector) could impact competition. Moving to more closed business models can limit competition – but it can also be a marker of strong dynamic competition: helping create efficiencies, lower prices, and diffuse, disseminate and encourage take-up of innovative AI services. This calls for a fact-specific, case-by-case analysis rather than a presumption than ‘openness’ is always the most pro-competitive outcome.

This issue paper will be followed by a subsequent report, which will launch in advance of CERRE’s upcoming AI and the Future of Competition event. For more on the impact of AI on digital market regulation, see our issue paper on AI Agents and Ecosystems Contestability.