> No wonder Tesla had to build a new substation for Cortex 2

They should be building entire power plants on-site. If they can't build a power plant at a given site, then build the whole thing, power plant and battery storage included, somewhere else.

An alternative to on-site power is a power plant that's close by (within tens of miles) then build dedicated power-transmission lines from the power plant to the data center.

Both solutions avoid the "we are going to bust the electrical grid with our power demands" problem.

> No wonder Tesla had to build a new substation for Cortex 2.

Err datacentres have literally always needed their own dedicated substations. The bar required for building substations is actually ludicrously low. Heck some commercial buildings without datacentres need their own substations. You pulling more than 1-2MW then having a dedicated substation is almost a necessity, unless you're feeding that power into a single load at the originally supplied voltage.

That's not to diminish your astonishment. No doubt Cortex 2 will draw a shitton of power, but the word substati

If you run computers underground they don't make heat? What if they are below sea level? Datacenters underground in Death Valley FTW!

What are you thinking? You can't just magic the heat away by building underground

Random internet commenter incorrectly assumes that they've discovered a silver bullet in their five-minute analysis that the experts have simply missed.

Must be a day ending in Y.

Underground? You'll have to pump air in there or your staff will die.

And you'll need to send the waste heat somewhere, maybe into the walls but the square-cube law indicates that this is not scalable. A really big data center will need to pump waste heat out through a medium, liquid or gas. Water is the obvious choice. So your underground data center ought to have a hyperboloid cooling tower sitting above it. Seems kind of pointless to have dug that big hole after all that.

I've never seen a more Dunning-Kruger post in my entire history on Slashdot, than what you just wrote there. You've in 4 sentences called engineers incompetent while demonstrating you don't know how cooling works, and proposed a solution that violates the basic physics behind conservation of energy.

Additionally you claim engineers forgot about weight when they literally didn't, and this article points out that weight very much has always been a fundamental consideration.

Please go home and sober up before po

How does putting it underground mean less heat? It gets hotter the lower you go, and on top of that the problem is heat generated by the servers themselves - where's it supposed to go in a room that's that well insulated?

If you'd had suggested building them atop mountains, you might have been on to something, but those places are typically hard to reach and thus construct upon, and you'd also still need to power them and, of course, connect them to the Internet.

One could use heat exchangers such that the cooling water and the sea water never intermix.

Consider it an investment in UFO science.

Usually the raised floor contains HVAC / cable runs. When I worked at rackshack back in the day, we had cooling towers that we could add that plugged into the floor panels. so... remove the floor and you would basically have to redo the entire datacenter.. cheaper to build a new one probablly. Still, I imagine they could find hardpoints and build up a frame, or use some kind of crane / suspension solution if they were locked into that location