Empire of AI Audiobook Review — Important Reporting, Too Much Backstory
A deeply researched, well-argued book that’s simply too long. The core chapters about the labor and environmental impact of AI are excellent. The rest is a very in-depth OpenAI history.
My Rating: ★★★☆☆ (3/5 stars: Good)
- Author: Karen Hao
- Category: AI, Business, Technology
- Published: 2025
- Runtime: 18 hours
This opens with a sharp, well-organized thesis: AI isn’t just a technology story, it’s a power story, and right now that power is concentrated in ways that are causing real, immediate harm to workers, communities, and the environment. That argument is thoughtful and well-evidenced. The problem is that it’s buried inside a blow-by-blow account of OpenAI’s internal politics that drags on.
I work in AI, so I’m reading and reviewing as many AI books as I can these days. What sets this apart from most AI books is the focus on immediate, measurable harm rather than long-term theoretical risk. That said, with all the headlines out there about the politics of OpenAI and other AI company leadership, I’m not sure that such an in-depth account of OpenAI actions is useful for all but the most interested readers.
The most important chapters are the ones on data labor, data centers, and the environment. Hao describes how AI companies rely on armies of contractors in the Global South to manually label and clean training data — often for wages that drop over time with no recourse — and in some cases, to review deeply disturbing content for moderation. One manager she quotes says the content is much worse in gen AI than in other industries. It’s a vivid, sobering portrait, and it’s the kind of reporting you won’t find in most AI books.
The data center chapter is similarly focused on environmental impact: one ChatGPT prompt requires roughly ten times the resources of a Google search, and cooling the servers that AI requires consumes valuable drinking water in communities that can least afford to lose it. These chapters are what the whole book could have been.
Hao’s closing argument is that AI governed by a handful of powerful companies becomes an empire. That’s reasonable, but it’s also true of most technologies that benefit from scale and network effects. Hao’s depth of reporting, based on extensive travel and research, makes her specific insights about AI’s impact powerful, but I felt that the closing wasn’t quite at that level.
The rest is a meticulous inside account of OpenAI: the founding, the Altman-Musk power struggle, Timnit Gebru’s departure from Google, Altman’s board firing and reinstatement, the Scarlett Johansson voice controversy, employment clawback clauses, and so on. It’s all well-sourced, and perhaps useful for someone who wants depth here, but it’s not for everyone. Unfortunately, the key labor and environmental insights were interspersed into this company history, rather than being a separate section that’s easy to skip to.
Put It To Work
- The labor behind AI is a human concern – and a supply chain risk: Every major AI product depends on low-paid, high-stress human labor that’s largely invisible.
- Environmental cost is measurable, not theoretical: Data centers require enormous energy and consume drinking water at scale. As AI usage expands inside organizations, those costs scale with it at both local and global levels.
- Compute is the real competitive moat: More than talent or algorithms, access to compute at scale is what separates the major players. Understanding that helps explain industry consolidation, the Microsoft-OpenAI relationship, and why chip export restrictions to China matter. Beyond looking at the AI labs, it’s worth keeping an eye on the companies that control their access to compute.
- The “mission-driven” structure is a power tool: A vague but compelling mission rallies workers, centralizes resources, and can be quietly reframed over time. Hao makes this point sharply near the end. Worth keeping in mind when evaluating any company whose governance relies heavily on stated values rather than legally-binding structures.
The Audiobook Experience
★★★☆☆
Author-narrated, which brings natural emphasis and inflection, but over 18 hours, the limitations of non-professional narration add up. Her voice occasionally sounds tired and her cadence gets repetitive in longer stretches. Nothing actively bad, just not outstanding.
Moderate multitasking potential. The OpenAI history sections are easy background listening; the chapters on labor and data centers reward more attention.
Audio or print? Either works. If you want to read this cover-to-cover, pick whichever you’ll get through easily. Chapters are clearly labeled in the menu, which helps with navigation in an audiobook this long.
Read It or Skip It?
Read it if: you’re specifically researching AI’s labor and environmental impact, or you want the most detailed account available of OpenAI’s internal politics.
Skip it if: you’re not specifically here for OpenAI’s internal history, since it gets in the way of the more valuable topics.
Related: The Optimist by Keach Hagey for a more neutral, efficiently paced version of the OpenAI story; The Coming Wave by Mustafa Suleyman for a more forward-looking take on AI’s risks and what governance could actually look like; The Thinking Machine by Stephen Witt for a broader and more insightful lens on tech industry leadership.
Listen Now
I only recommend audiobooks and resources I’ve personally experienced. This post contains affiliate links — if you purchase through them, I earn a small commission at no extra cost to you.
