The Entire AI Data Center Explained — From Electricity to ChatGPT
The short version: Leo Cui's walkthrough of the AI 'token factory' is a solid technical primer, but the real story is who pays for all that electricity once the GPUs light up.
Our Take
Credit to Leo Cui for laying out the AI pipeline in plain terms — fiber, GPUs, high-bandwidth memory, cooling, and the training-versus-inference split that explains why a single chat reply now requires small-city levels of power. It's a good technical primer, and it makes something clear that our readers already feel in their utility bills: the 'AI factory' isn't just silicon, it's substations, transmission lines, and water-cooling loops sited in someone's backyard.
What videos like this tend to skip is the local paperwork behind that electricity. Every gigawatt of GPU capacity Leo describes got approved somewhere — by a county commission, a utility regulator, or a state economic development office offering tax abatements to lure the build. Those deals rarely get the same explainer treatment as the tech stack, even though they decide whose rates go up, whose water table gets tapped, and whose farmland turns into a substation buffer zone.
If this video made you curious about the hardware, let it also make you curious about the contracts. Check our facility map to see if one of these token factories is being built near you, and visit our take-action page for ways to push for transparency before the concrete gets poured.
This is GridWatch the USA’s original commentary. The video above is the work of Leo Cui, Ph.D., CFA, published on YouTube — full credit to the creator.