Empire of AI: Dreams and Nightmares in Sam Altman's OpenAI

"Empire of AI: Dreams and Nightmares in Sam Altman’s OpenAI by Karen Hao is a deeply reported account of how modern AI—especially generative AI—has actually been built, and what that reveals about power in the tech industry. At its core, the book argues that AI is not just a technological breakthrough but the foundation of a new kind of empire, shaped by a small number of companies with vast access to data, compute, and capital. Focusing on OpenAI and its CEO Sam Altman, Hao shows how ideals about openness and safety have collided with commercial pressures, geopolitical competition, and the sheer cost of building frontier models. One of the book’s key insights is that the AI boom depends on hidden infrastructures and labour—from energy-hungry data centres to the often-overlooked human work of data labelling and content moderation. This challenges the sleek narrative of AI as purely digital or autonomous, revealing it instead as a messy, global system with real-world consequences. Hao also traces how control over AI is becoming increasingly centralised, raising questions about accountability, governance, and who ultimately benefits. The ‘nightmare’ side of the title points to risks like concentration of power, lack of transparency, and the potential for misuse at scale. I chose this book because it provides something many AI titles don’t: serious investigative depth. It grounds the discussion in reporting rather than speculation, and gives readers a clear-eyed view of the institutions shaping AI—making it an essential counterbalance to more optimistic or abstract accounts. Yes—that’s very much the concern Karen Hao raises, though she presents it more as a structural tendency than an inevitability. Her argument is that modern AI has unusually strong winner-takes-most dynamics built into it. Training and deploying frontier models requires vast amounts of capital, data, specialised talent, and computing infrastructure—resources that are already concentrated in a small number of companies. That creates high barriers to entry, making it easier for a few dominant players to pull further ahead, much as we saw with earlier tech platforms, but potentially on a larger scale. However, the book doesn’t claim we’ll literally end up with ‘a handful of trillionaires controlling everything.’ The more precise worry is that power over key AI systems—and therefore over information, labour, and decision-making—could become highly centralised in a small cluster of firms and their leaders. That concentration could shape markets, public discourse, and even geopolitics. At the same time, there are countervailing forces. Governments are beginning to regulate AI, open-source models are lowering some barriers, and competition—especially between the US, China, and others—may prevent a single monopoly from emerging. So the trajectory isn’t fixed. The useful way to frame Hao’s point is: AI is likely to amplify existing concentrations of power unless actively checked—and whether it leads to extreme inequality or a more distributed ecosystem depends on policy, competition, and how the technology evolves."