LISBON, February 20, 2026 — The Portuguese Competition Authority (AdC) has published a new short paper analysing competition issues associated with access to chips used in the training and execution of Artificial Intelligence (AI) models.
According to the AdC, access to AI chips, as well as related hardware, particularly through cloud computing services, is a key factor in the development of AI systems. However, the authority notes that the value chain for these chips has structural characteristics that make it prone to high levels of concentration.
The AdC highlights that AI chip production relies on a complex global value chain, with both the design and manufacturing phases exhibiting a high degree of concentration. This situation stems from structural factors such as significant economies of scale, high investment costs and technological complexity. In this context, constraints may arise along the value chain, leading to supply rigidity and increased prices for AI chips. These factors may reinforce barriers to entry and expansion, particularly for startups and AI vendors seeking to train and deploy advanced models.
The authority also points to growing vertical integration in the sector. Some leading chip designers and cloud service providers have expanded along the AI value chain, developing integrated ecosystems that combine hardware, software and services. At the same time, strategic partnerships and cross-investments have been established between sector operators.
According to the AdC, these dynamics may raise competition concerns, including the granting of privileged access to sensitive technical and commercial information, the alignment of incentives among operators, reinforcement of technological dependence (lock-in), reduced interoperability, and an increased risk of excluding competitors or leveraging market power into adjacent segments. The document cites CUDA, NVIDIA’s development platform, as an example whose widespread adoption and associated network effects strengthen the company’s position in the AI chip market.
The AdC further notes that access to public high-performance computing (HPC) infrastructure may help mitigate some of the identified bottlenecks. However, its pro-competitive impact will depend on transparent, objective and non-discriminatory access criteria.
With this publication, the AdC continues its analysis of the competitive impacts of generative AI, following previous short papers on data access and use, the opening of AI models and AI labour markets, as well as an Issues Paper published in 2023. The authority stated that this series of documents seeks to identify emerging risks to competition in digital markets and to promote preventive and informed action in line with its mandate.
