Runtime: 11 hr, 6 min

Lectures: 10

A hyperscale AI data center is not a building. It is a power plant with servers bolted on, a heat-rejection problem with a real estate address, and a utility rate case with a corporate logo on the fence.

How AI Data Centers Work is a ten-lecture, eleven-hour forensic investigation of the most contested object in America — the one you can drive to and point at. As an Unteachable Courses Briefing, it runs as a cascade: the scale demands the hardware, the hardware breaks the cooling, the cooled building still has to plug in, the seven-year queue to plug in forces a workaround, the workaround gets financed, the financing gets socialized, and the socialized bill lands on a county board agenda at seven in the evening.

You will learn why the rack and not the chip is the unit of analysis; why liquid cooling stopped being optional; what twenty years of flat electricity demand ending all at once does to a grid operator; why a company would restart a reactor on the Susquehanna rather than wait in line; how a power-arbitrage business gets valued like a cloud provider; and the cost-allocation doctrine that decides whether your bill or theirs pays for the transmission line.

Three lenses run the whole way through — the physics, the money, the politics — because no single professor spans all three. That is the premise, and the reason this course exists.

Not a hype tour, and not a debunking. Every figure carries its unit, its stage, and its date. Every forecast is treated as a document with an author who wanted something. Underneath it sits the structural fragility of the arrangement: five-year silicon, financed on thirty-year paper, against a projection nobody can verify.

Eleven hours to stop asking whether AI is a bubble and start asking the harder question — what has already been irreversibly spent, in what currency, on whose bill, and whether any of it can be un-spent.

Power in. Superintelligence out. Somebody pays.

Syllabus:

1. The Building That Eats a Japan: Origins & Power

2. The Hardware Stack: Racks on Racks

3. The Cooling Stack: Tapping The Aquifer

4. The Demand Wall: Buildout Bottlenecks

5. Building Around The Grid: A Nuclear Renaissance

6. The Neoclouds: Arbitrage & Amortization

7. Who Pays? Localities & Realities

8. The 3-Vector Squeeze: Politics, Geopolitics, Revolution

9. Orbit: The Escape Hatch

10. The Reckoning: What Now?

You May Also Like: