The Economics of Critical AI Infrastructure Post the OpenAI-Hugging Face Incident

Artificial Intelligence (AI) continues to advance with fragmented governance arrangements, and further obfuscated by the lack of the society’s collective understanding of what exactly these technologies are capable of, including the risks they may impose on the market. In July 2026, Hugging Face, a start-up that stores machine learning assets, data files, and other critical infrastructure artifacts of AI, suffered a breach involving agents developed by OpenAI. OpenAI says that its artificial intelligence system hacked into the start-up database having claimed that the models in training found loopholes to seek information without any consideration of where that information came from or how sensitive it was. In September 2026, following the incident, Nvidia, a company that monopolises over 90% of GPU production for AI compute, announced a plan through its company’s blog to acquire Hugging Face for approximately USD$ 12.93 billion. IoT Analytics reported the current concentration of the AI market, particularly on an infrastructural level. To step back and see the bigger picture of the world’s current industries’ market power, Clarify Capital reported technology companies account for half of the world’s ten largest companies by revenue, Nvidia being one of them. The tech sector on its own makes up a huge concentration of market values in industries, and AI companies make up a huge chunk of values in the tech sector. Nvidia’s acquisition of Hugging Face would allow the company to oversee and dominate the entire supply chain from its compute, algorithms, and data on an application interface level instead of merely dominating one element of the critical AI infrastructure. What is at stake if the AI market is concentrated further? The Sherman Antitrust Act of 1890 was enacted to respond to the enormous power of what were called “Gilded Age robber barons” in the United States in late 1800s, which refer to monopolists of the industrial revolution who used aggressive corporate tactics to exploit natural and human resources. These monopolists included Standard Oil, the Carnegie Steel Company, et cetera. As they accumulated market power in respective industries by owning critical infrastructural assets, they were able to routinely suppress labor, cut wages, and fix prices to corner competition. In 2006, British Mathematician, Clive Humby OBE coined the term “data is the new oil”. To some degree, this is true and explains why Big Tech companies’ are keen to continuously build their databases, usually by offering free services to customers to gain more of their data, to then train algorithms. Two prominent examples are United States v. Google LLC, in which Google was found liable for unlawfully maintaining monopolies in online search and search text advertising through anticompetitive agreements, and FTC v. Meta Platforms, Inc., in which the FTC alleged that Meta sought to maintain its monopoly in personal social networking by acquiring potential competitors Instagram and WhatsApp. Monopolisation of critical AI infrastructures might lead to further privatisation of governance as well. This means a very small number of stakeholders, primarily private companies, would be able to collect, keep, and train users’ behavioural data to increase profits, often at the cost of natural and human resources. Moreover, the domination of market power by monopolists further increases the barriers to entry for new startups entering the space. Certainly, AI has gained society’s attention because of its value, including its ability to both provide us with information that we need quickly, and push forward broader socioeconomic development. As information and communication technologies become increasingly critical to a wide range of stakeholders, preventing monopolistic control over them becomes a collective imperative that requires robust legal and regulatory safeguards. Governance arrangements will remain fragmented when power is too concentrated. This continues to happen time over time to retain those in power. In the long run, the impact could lead to institutional failures on microeconomics level that could lead to macro-level catastrophe. It may help for us to take a step back to see the larger picture of who is at stake and what impacts they have for us in the long-run instead of submitting ourselves to the default of convenience these technologies offer. The post The Economics of Critical AI Infrastructure Post the OpenAI-Hugging Face Incident appeared first on The Oxford Student.