Hacker News Reader

Comment by petre | Hacker News Reader

The heat has to go somewhere and that is the environment. 300 MWh is enough energy to boil over 3k metric tons of water. That's 107 medium fuel trucks for perspective.

View on HN

300 MW hr is approximately nothing to the broader environment. A constant gigawatt load is (off the top of my head, probably off a bit) something like 5 sq km of solar over a 24 hour period on average. Granted some of that light would otherwise be reflected but that gets us in the rough ballpark.

In local terms its a fair bit of heating but zooming out it's a drop in the bucket.

This seemed high to me.

According to Google, one ton of water takes about 730kWh to boil. So I think you’re off by an order of magnitude, it’s only ~450 metric tons.

(But this assumes that no heat is radiated away in other forms.)

Google confused you. One needs 730 kWh to fully evaporate 1 ton of water.

Otherwise it's 1.16 Wh/kg (or kWh/ton) to rise the temprature by 1°C. Thus one needs a delta of 80°C, so 93 Wh to boil a kg of 20°C water. That's what my napkin math was based on. I used that metric a lot to calculate heat deltas in storage tanks.

I think it's the context that's confusing here. Given the topic the first thought is evaporative cooling but IIUC your intention was to give perspective by comparing to a volume of water raised to the boiling point.

This makes it seem like you think all energy consumption leads to water boiling? At the peak of a sunny day in the american southwest, a random square mile receives over 2000 megawatts of power from the sun. In a 24-hour period that same square receives ~16 gigawatt-hours of solar energy. It doesn't all get used to boil water. And as others have said, with a closed loop cooling system, no water is evaporated that isn't also re-condensed.

My 1500W space heater could boil 4.32 gallons of water every hour. But it isn't.

Luckily AI data centers produce nowhere near that amount of heat. Remember the heat is waste and 300 MWh is the total draw. Some of that energy becomes heat. That ratio is somewhere like 100:1 though. Also, the waste water is only like 10F hotter than the intake. We build GW sized PP all the time and they will leak far more heat (as like on the order of 100x) than a 300 MWh AI data center. Thought there were supposed to be engineers on this site.

> Some of that energy becomes heat.

I'm neither for nor against, but on the physics here: basically all of the energy input as electricity is transformed to heat leaving a datacenter. Only a tiny tiny fraction is emitted as radiation (eg floodlights outside or light in fiber optics) or as kinetic energy (air moving away from fans/vents).

Computers are machines for turning electric energy into heat energy, plus some small useful side effects.

Also powerplants are quite (relatively) efficient in terms of heat-to-energy output, often >50% afair. So a 1GW power plant will generate something like 2GW of heat (or less), not 30GW.

Some? Where else does the energy go?

An electronic circuit drawing 1W of input power will dissipate all of that 1W as heat (assuming it's not outputting light or sound or other physical side effects).

In the macro physics sense, all energy eventually becomes heat somewhere. My 1500W space heater is perfectly efficient and produces 1500W of heat, and my 1500W crypto mining rig does too. Thermodynamically there's no difference. Where would the energy go if not into heat? Even energy that does things like push air or emit radiation, eventually becomes heat somewhere. In the case of a power plant, that somewhere is just very far away with the end user.

Though there might be practical differences though between the excess heat intentionally exhausted, and other heat. Just speaking from a very macro sense.

Yeah it gets radiated into space

Reply on news.ycombinator.com