Missed the edit: So a single wafer may have a thousand die on it. Wafers are processed in bundles of 25. But each one of those die is just one component. In other words, all the die on a wafer and in the bundle are going of the exact same component.
So an entire bundle (really called a lot) may just be an IC that regulates the power supply or something.
I have no doubt that the computer manufacturers have quantified it.
I’d be satisfied with order of magnitude. Is it around 10 gallons per computer? 1000?
According to the probably-not-stupid people at Stanford, 10 gallons per finished chip. I gave up counting the number of chips in my computer at somewhere over 20, although a lot of them are very small. But some of them, though small, are pretty sophisticated and probably needed the whole 10 gallons.
Fer example: CPU: 1 seriously large chip. GPU: also, 1 seriously large chip. Motherboard chipset: usually 2-3 largish chips. GPU support chips: 2 or 3 as well. Multiple on-board NVRAM or EEPROM chips. Various low-level support chips. I/O controllers. Memory, usually 4-8 per DIMM. At least 1-2 chips per hard drive or optical drive. A chip in the keyboard. A chip in the mouse. A chip or two in the display, if it’s a modern flatscreen.
I’d be surprised if a desktop system didn’t require hundreds of gallons of water for its semiconductors.
Environmental people do what’s called “lifecycle analyses” that attempt to quantify the resources used in a product’s lifecycle from manufacture to disposal. This paper (payment or academic credentials required) cites a 1993 analysis by US manufacturers saying that production of a desktop computer + CRT monitor uses 27,700 kg of water (~7300 US gallons). It also says… you know what, I’ll just quote it:
So it’s not easy even for the pros. I guess you can assume “several thousand gallons”, absent better data.
Interesting, but it’s not passing the sniff test, most specifically because of the item number 4. When you scale it up to actual production numbers the required volume just doesn’t work out. There’s still data missing. How much of that 7300 gallons is recycled into producing the next workstation off the line?
Now I’ve got go see what I can dig up. Fine can of worms I opened up…
It is possible (my guess) that the 7300 figure also includes water used for hydroelectric power, mining and cleaning the materials for manufacture, etc. In a LCA, it’s not just “How much water is used by the Foxconn factory in China?” but how much water is used by all the suppliers leading up to that point as well – excluding everyday human activities such as drinking and bathing for the workers, obviously. How far an individual LCA goes depends on the actual paper.
If you do manage to find better, more specific data, please do share. I’m very curious too 