OpenAI Plans $1.4 Trillion Compute Buildout with AWS

The Significance of One Gigawatt

One gigawatt is an enormous amount of power.

To put it in perspective:

1. A Large Nuclear Reactor: A single typical nuclear power reactor has an output of about 1 gigawatt.
2. Peak Power of 1.3 Million Homes: A gigawatt is enough to power approximately 750,000 to 1.3 million average American homes, depending on the region and time of year.
3. 1.3 Million Horses: It's equivalent to about 1.3 million horsepower.
4. Lightning Bolt: A single lightning bolt at its peak can deliver about 1 gigawatt, but only for a few microseconds.

The Truly Mind-Boggling Part: "Every Week"

Adding a gigawatt of capacity is a massive undertaking for a country, let alone a single company. Doing it every week is an unprecedented and almost unimaginable scale of infrastructure expansion.

Let's do the math:

• Per Year: 1 GW/week * 52 weeks = 52 Gigawatts per year.
• Comparison to the Entire UK: The United Kingdom's total electricity generation capacity from all sources (gas, nuclear, wind, solar, etc.) is roughly 75 Gigawatts. This initiative would be adding two-thirds of the UK's entire national grid capacity every single year.
• Comparison to a Massive Dam: The Three Gorges Dam in China, the world's largest power station by capacity, is about 22.5 Gigawatts. This project would be adding the equivalent of two Three Gorges Dams every year.

Why Does AI Need This Much Power?

This power isn't for lighting offices; it's to run the tens of thousands of high-performance computer chips (GPUs) that train and run massive AI models.

• Energy-Intensive Computation: Training a single advanced AI model can consume more electricity than 100 homes use in an entire year.
• Constant Operation: These data centers run 24/7 at full throttle.
• Heat Generation: All that power is converted into heat, requiring equally massive and power-hungry cooling systems.