OECD: Artificial Intelligence Does Not Yet Automatically Increase Corporate Market Power

A new OECD study does not support the simple claim that using artificial intelligence automatically increases corporate market power. Data from French and Portuguese companies between 2011 and 2022 show that users of earlier forms of AI tended to be larger and more productive, yet their mark-ups were not consistently higher. The result matters because adoption of the technology alone does not prove weaker competition or abuse of a dominant position. Technology ownership presents a different picture. In sectors where AI-related patents are concentrated among a small number of owners, the study also finds greater market concentration. Mark-ups grew faster among information and communications companies that owned such patents. The authors therefore describe the environment as “dynamic but uneven”. Not every user of a model gains the same advantage; data, skilled workers, capital and the ability to integrate the system into operations are decisive. Generative models may reduce entry costs for smaller and younger companies. They can help create software, product designs, marketing material or customer support without a large team. At the same time, they may widen the gap between businesses with complementary assets and those that merely pay for access to a model. If a few companies control the key data, computing power and distribution channels, a cheaper tool will not by itself open the market. The OECD also highlights acquisitions of start-ups by established companies. Some deals give a young business capital and access to global markets. Others may remove a future competitor before it threatens the buyer's main product. The study uses the term “killer acquisitions” for cases in which the buyer purchases an innovation not to develop it, but to stop it. Competition authorities often find it difficult to prove that intention in time. Regulators should therefore look beyond current market shares. They need data on patents, access to computing capacity, contracts with cloud providers, movement of specialist workers and purchases of start-ups. When reviewing a merger, control over future technology may matter more than the target company's present revenue. Policies supporting small businesses should also fund skills and work with proprietary data, rather than only vouchers for the use of ready-made models. The study has an important limitation. Most company data ends before the mass adoption of generative AI, and the detailed analysis covers only two countries. The relationships identified do not by themselves prove causation. The OECD therefore recommends continuous monitoring. The practical conclusion for company leaders is restrained: technology may raise productivity, but lasting advantage emerges only when a model is combined with unique data, expertise, distribution and customer trust.

OECD: Artificial Intelligence Does Not Yet Automatically Increase Corporate Market Power

The full article is available to users registered at Hard Skills