The Age of Models: How Artificial Intelligence Became Infrastructure — and Who Now Controls It. (Infrastructure - The Hidden World)

In April 2026, the United States Treasury Secretary and the Federal Reserve Chair convened America's largest banks for an emergency briefing. The reason was not a market crisis. It was the announcement, that same morning, of an artificial-intelligence system capable of finding software vulnerabilities at industrial scale. The same week, the Wall Street Journal reported missed revenue targets at OpenAI; SoftBank fell ten per cent; the French government issued a directive on European digital sovereignty; and Nvidia became the worst-performing share of America's largest technology firms. The Age of Models is the story of how artificial intelligence stopped being a consumer product and became infrastructure — the kind of infrastructure on which national security, commercial life, and ordinary daily routines now silently depend. It is part of an ongoing series about the technologies that underpin the modern world: a series that has already taken up wireless networks, undersea cables, and the data centres in which the cloud actually lives, and that argues, in each case, that what is presented in public discussion as a neutral technical system is in fact a political artefact, shaped by particular parties with particular interests. The book makes three central claims. First, that artificial intelligence has, over the past four years, become infrastructure in the strict sense: essential, durable, externality-producing, and a matter of state policy rather than ordinary commerce. Second, that the industry producing this infrastructure is among the most concentrated of any general-purpose technology in modern history — at the chip layer, the data-centre layer, the model layer, the capital layer, and the regulatory layer above. Third, that the question of who controls this infrastructure has not been settled, and is being contested in real time by laboratories, hyperscalers, capital providers, governments, and the publics affected by their decisions. The frame is summarised in three words: capability, concentration, control. Across twenty-four chapters, the book traces: Capability — what these systems can actually do, where they fail, and what the rise of reasoning models and AI agents means in practiceConcentration — the small set of laboratories, chip designers, foundries, hyperscalers, and capital providers that produce frontier AI, and the chokepoints at every layer of the stackControl — who decides what is released and to whom, how governance is being constructed, and the late-April 2026 events that have begun to test the structureSpecific events covered include the Stargate announcement, the DeepSeek-R1 release, the OpenAI–Microsoft renegotiations, China's parallel AI ecosystem, the EU AI Act, the rise of multi-agent systems, the limits of current scaling, and the open-versus-closed contest that has shaped the industry's politics. Written for the intelligent general reader, the book is detailed enough to repay careful attention, plain enough to be read without specialist preparation, and honest enough to acknowledge where its observations are tentative. Twenty-four chapters. Four appendices, including a working glossary of central concepts and full profiles of the principal companies and organisations. The frame for understanding the technology that is reshaping the world. Read more