Empire of AI

Empire of AI Summary

Dreams and Nightmares in Sam Altman's OpenAI

by Karen Hao

  • 14 min read
  • Published 2025
  • 9 takeaways

The chat box feels weightless. Karen Hao follows the wires into the furnace room, the back office, and the boardroom, where “intelligence” starts to look less like magic and more like empire with better branding.

What you'll learn
  • Why AI needs an empire
  • How scale became power
  • The workers beneath automation
  • What data extraction really buys
  • Who holds the emergency key

Key point 1

The chandelier hums

The demo looks clean because the mess has been moved out of sight.

Karen Hao, a technology journalist who has reported from inside and around the AI industry, treats artificial intelligence less as a magic trick than as a built place. There is a bright hall where users ask clever questions. There are service tunnels where data, money, labor, power, and politics keep the lights on.

Her core claim is sharp: the leading AI companies did not merely discover a new technology. They built an empire around it, with borders drawn by compute, data access, and control over the story of the future.

OpenAI is Hao’s main palace because its public mission promised safe AI for everyone while its business model pulled it toward speed, secrecy, and market power. The book asks who pays for intelligence when intelligence becomes an industry.

Key point 2

The charity that learned to charge admission

In 2015, Sam Altman, Elon Musk, Ilya Sutskever, and others helped launch OpenAI as a nonprofit with a grand promise: build artificial general intelligence, often called AGI, and make sure it benefits humanity.

That word, humanity, does heroic work. It sounds like a public square. It can also hide who gets a key.

Hao’s point is that OpenAI’s origin story matters because the company wrapped itself in moral purpose before it became a market force. The nonprofit form was meant to stop a dangerous technology from being captured by one company or one state. Yet the task demanded huge amounts of money, rare talent, and computing power. The public hall needed private wiring.

A mission can make a company look smaller than the power it is gathering.

By 2019, OpenAI had created a capped-profit arm so it could raise serious capital while still claiming a public-interest mission. The structure was clever, but clever structures are still structures. They do not cancel pressure from investors, partners, employees, and rivals.

A mission statement is a poor seat belt when the car is built to go faster every month.

This matters beyond OpenAI because many powerful technologies now arrive dressed as rescue missions. The language is warm, but the incentives are cold enough to store fish. Hao asks readers to follow the incentives rather than the speeches. Once the company needs scale to survive, the palace stops being a public gift and becomes a place with guards, sponsors, and rooms most people will never enter.

Key takeaways

Key point 3

The furnace became the strategy

Key point 4

The servants under the marble

Key point 5

A harvest with lawyers at the fence

Key point 6

Who holds the emergency key

Key point 7

When smaller fires changed the map

Key point 8

The palace becomes a city plan

Key point 9

Try this

Continue reading the full book summary and unlock all remaining key takeaways.

Get full summary

About the author

Karen Hao

Karen Hao is an award-winning technology journalist whose reporting has taken her inside the companies, labs, and political machinery shaping artificial intelligence. Formerly a senior AI reporter at MIT Technology Review and a contributor to major outlets including The Atlantic, she brings rare fluency in both the technical architecture and the human cost of the industry’s favorite miracle machine.

Related topics

Want to keep reading this summary?

Get full access to complete summaries and audio versions in one place.

Continue to onboarding

Related books

Keep learning with similar reads

Unlock full library

Frequently asked questions