Quick Take
- Narration: Karen Hao reads her own book with a journalist’s controlled intensity, restraint in the delivery makes the documented evidence more devastating than dramatization ever could.
- Themes: OpenAI’s internal culture, global resource extraction, power consolidation in the AI industry
- Mood: Investigative and unflinching, with the mounting pressure of a long-form exposé
- Verdict: The definitive account of how OpenAI became what it is, Hao’s on-the-ground reporting across Silicon Valley, Kenya, and Chile produces an essential and troubling book that no adjacent AI title can replace.
I cleared a full weekend for Empire of AI. At 17 hours and 50 minutes, Karen Hao’s investigation into OpenAI and the broader AI industry demands that kind of sustained attention, and it earns it across every one of those hours. This is not a book you listen to in commute fragments. It is a book that requires you to sit with what it is showing you.
Hao is a longtime AI reporter who covered OpenAI from 2019 onward, developing the deep sourcing inside the organization that makes this account something different from every other AI book I have read. She was not reconstructing OpenAI’s history from interviews conducted after the fact. She was in contact with the story as it happened, including the November 2023 events surrounding Sam Altman’s firing and return, which she covers here in more detail than has appeared anywhere else. The behind-the-scenes account of that episode alone is worth the listening time for anyone trying to understand who is actually running the most consequential technology company in the world and why they make the decisions they do.
The Colonial Frame and Why It Holds
The book’s most distinctive intellectual move is the sustained analogy between the AI industry’s resource extraction model and the structural patterns of historical colonialism. Hao is not making this argument rhetorically. She is making it empirically, through interwoven reporting from three locations: Silicon Valley, where the engineering decisions are made; Nairobi, where data labeling happens for subsistence wages; and Chile, where the water consumption of data centers is measurably depleting local water tables.
One reviewer describes this frame as having an agenda, and they are right that Hao does not pretend to neutrality. But the reporting itself is not tendentious, it is documented. The conditions she describes in OpenAI’s Kenyan outsourcing operations were subsequently confirmed by multiple other news organizations. The water-usage data for data center infrastructure is publicly available. The agenda is in the choice to juxtapose these three vantage points rather than keep them separate, and that juxtaposition is precisely what makes the book valuable. The AI industry narrative tends to be told from Silicon Valley outward. Hao tells it from the extraction sites inward.
Altman’s Firing and What It Actually Revealed
The sequence of events surrounding Sam Altman’s November 2023 removal by OpenAI’s board, and his return five days later with greater power than he had before, has been written about extensively. Hao’s account is the most complete. Her access to sources inside the board, the technical staff, and Altman’s inner circle gives her a picture of what was actually at stake that the public reporting missed or misframed. The episode was not primarily a governance dispute or a safety disagreement in her telling. It was a confrontation between two different theories of what OpenAI was supposed to be, and the resolution of that confrontation made clear which theory had won.
This sequence is among the most compelling investigative writing in recent technology journalism. TIME Magazine recognized Hao as part of TIME100 AI 2025, specifically crediting Empire of AI with fundamentally shaping people’s perceptions and understanding of OpenAI globally. That is an unusual degree of institutional acknowledgment for a single piece of technology reporting.
The Porchlight Award and What the Recognition Signals
The book is a National Book Critics Circle Award finalist, a New York Times Notable Book, and the Porchlight 2025 Business Book Award winner. Those three recognitions together tell you something about the book’s unusual range: it is being read by literary critics, general readers, and the business audience most directly implicated in its arguments. Vulture called it startling and intensely researched. Tim Wu in the New York Times called the reporting excellent and deeply reported. The depth is real, and the listening experience reflects it.
Duration, Scope, and the Weight of What Is Here
Eighteen hours is a significant commitment, and it is worth being clear about what the runtime reflects. This is not a padded book. Hao has three interwoven narratives, the internal OpenAI story, the global extraction story, and the intellectual history of the AI nonprofit-to-commercial transition, and she gives each the space it needs without rushing the connections between them. Hao narrates with the controlled delivery of an investigative journalist presenting evidence. She is not enjoying this material. She is documenting it. That restraint makes the more damning passages land with far greater force than any performed outrage would.
Frequently Asked Questions
Does the book require prior knowledge of OpenAI or the AI industry, or is it accessible to general readers?
Hao writes for a general audience and contextualizes the technical and organizational history from the beginning. Prior knowledge of OpenAI or AI development will deepen the reading experience, but is not required. The book is structured as journalism rather than technical analysis.
How does Hao’s account of Sam Altman’s firing and return in November 2023 differ from what has been reported elsewhere?
Hao’s account draws on deeper sourcing than most prior reporting, including perspectives from inside the board, the technical staff, and Altman’s close circle. Her framing of what the episode revealed about OpenAI’s power structure is more complete than the governance-dispute or safety-disagreement framings that dominated public coverage.
Is the colonial framing Hao uses for AI resource extraction a genuine analytical tool or polemical positioning?
One reviewer notes the book has an agenda while praising its depth and sourcing. Hao’s framing is documented through on-the-ground reporting in Kenya and Chile, not purely asserted. Whether you find the colonial analogy proportionate is a matter of interpretation, but the underlying facts she documents are verifiable and have been confirmed by other reporting.
Does Karen Hao’s self-narration work across the nearly 18-hour runtime, or does the length expose limitations in the performance?
Hao reads with the controlled delivery of an experienced journalist presenting evidence rather than a performer seeking to entertain. The restraint is consistent throughout, and reviewers do not raise pacing or performance as concerns. For material of this weight, the measured delivery is well-matched to the content.
