The Data Center
A building that is mostly power plant and plumbing, with computers inside.
How big?Football fields of floor space. The biggest AI campuses use as much power as a city.
Real: What it looks like.
Keys: arrows rotate · + / − zoom · 0 reset · 1–4 views · S signal · T tour · L labels · Space spin. Models are stylised and built from code: proportions are honest, details are simplified.
What it is
Inside are long halls of racks arranged in rows. Cold air or water comes in, and hot air is trapped in 'hot aisles' between rows so it can't mix back.
Outside is a small power utility: a substation steps down high-voltage grid power, battery rooms (UPS) bridge brief outages, and rows of diesel or gas generators wait for longer ones.
The cooling plant (chillers and cooling towers) carries the heat out. Fiber-optic cables connect it all to the internet.
Why AI needs it
Training frontier models needs tens of thousands of GPUs in one place, wired tightly. Finding enough electricity, water and land is now one of the biggest limits on AI.
Every labeled part
- 1
Data hall
Rows of racks. X-ray the walls and roof to look in.
- 2
Hot-aisle containment
Enclosed aisles keep hot exhaust air from mixing with cold supply air.
- 3
Roof + air handlers
Explode to lift the roof.
- 4
Cooling towers
Evaporate water to dump heat into the sky. Signal mode shows the vapor.
- 5
Chillers
Refrigeration machines that cool the water loop on hot days.
- 6
Substation
Transformers step down grid power. Yellow flow in Signal mode.
- 7
Batteries (UPS)
Keep everything running for the seconds it takes generators to start.
- 8
Backup generators
Rows of engines for long outages.
- 9
Fiber in
The data highway to the rest of the world.
Try it · concept lab
Where does the heat go?
Turn up the power, then change the cooling.
Chip temperature
97 °C
Status
throttling 🔥
For 100 MW of computers, the building draws 125 MW. That's 25 MW just for cooling, power conversion and lights.
Simplified model: temperature = room + power × thermal resistance. The resistances are illustrative.
Big idea: Nearly every watt becomes heat. Liquid carries it away far better than air, and every bit of building overhead costs real power.
Swap it: other ways to do the same job
Cloud region vs. owning one
Renting GPUs from a cloud is flexible; building your own is cheaper at huge scale but takes years.
Edge data center
Small sites close to users. Faster responses, tiny compared with a training campus.
Power sources
Grid, solar and wind contracts, batteries, gas turbines, and new nuclear deals each trade cost, cleanliness and reliability.
Go deeper
Who gets the power?
The biggest AI data centers can use as much electricity as a city. Towns that host them can get tax money and some jobs, though many sites are offered tax breaks to come. They can also strain the local power grid and water supply. Most run on a mix of grid power, gas, solar, wind and nuclear deals. Different towns have chosen differently, and neighbors often disagree.
Figures are rough, as of 2026.
Talk about it
- Q1
Most of this building is power and cooling gear. Why do the computers inside need so much help?
- Q2
If a data center came to your town, what would you want to know before deciding how you feel about it?
- Q3
A data center can bring tax money and some jobs, but needs lots of power and water. How would you weigh those?
For grown-ups: there are no right answers here. Ask a question, then ask "why do you think that?" The reasons matter more than the answer.
Printable question sheet (PDF)