Hacker News Reader

You’re right, fabs need more water than a swimming pool, maybe a big swimming pool.

The fabs in Arizona are likely not going to be the main cause of water shortages there. How do I know this? Cause when Motorola had its massive (and inefficient) fabs running there in the 90’s-00’s and there were water shortages, they kept running just fine off their on premise reserves.

Modern fabs are even more efficient with water usage.

I know this because I have visited the old fabs and they are quite good at re-using and capturing the water since it’s inside an air controlled bunker. Modern (and older) fabs aren’t open to air, so water loss is minimal.

The latest 300mm fabs use 10-50X the power of what the old 8" fabs used (I know this for a fact...). That power gets turned into heat and that heat has to be removed. I work at an "inefficient" 200mm factory that has pretty good internal recycling (~65%)...we still use 600K gallons per day.

A single state of the art EUV tool uses ~2MW of power...<0.1% of that gets to the wafer. Most of the rest of that is lost to heat...and that heat goes where?

Seeing those numbers, I wonder how much that affects the local microclimate, as that adds a lot of moisture and heat into the air. And if there are ways to add plant life in the surrounding areas that could benefit from it, and from which people can obtain a yield.

I'm not so sure about the hotter months, but during the colder months, it can potentially be used for greenhouses, or for when temperature dips during nighttimes. Lots of tropical plants want hotter, moist air and will die from frost.

An acre-foot is about 325K gallons, and an irrigated farm uses 1 or 2 acre-feet per acre per year. So the "inefficient" fab you work at uses less than 2 acre-feet/day, or maybe 700 per year. About what a 500-acre farm uses.

Arizona has about 1.3 million acres under irrigation. So 2,500 similar fabs would use all the irrigation water.

It's a lot of water, but maybe not a deal breaker?

The USDA says Arizona cotton farmers make 1,138lbs/acre/year (sorry world, not converting units this time). There are 165,500 acres growing cotton in Arizona, so the 500 we stop to get water for Intel is negligible.

So we are talking about agriculture forgoing 569,000 lbs of cotton per year. Farmers get about $0.60/lb for cotton. The impact on the agricultural industry is $341,000/year.

Buy out the worst cotton farmer in the state. Take his water. Move on.

Definitely not arguing that the heat doesn’t need to go somewhere. You’re absolutely right.

All I remember is family saying they had effective means for recycling the water without losing it in these boiler chambers.

I’m no expert, but they were saying significantly higher efficiency than 65%.

Assuming that those 600kgal start out ice cold and all end up evaporated at the boiling point, that’s 146MW of cooling. LCoE for plain solar power is about $31/MWh. Assuming you could discharge heat without evaporation at a COP of 4, that’s about $26k/day to buy power. Not at all cheap, but not insane. Interestingly, that is a lot more than the cost to buy that much water wholesale, even in Arizona, so I guess this is why people like cooling towers.

600k gallons for us is not all evaporated, thats just what we take in from the city and discharge back to them. There are temperature requirements for discharge as well so you can't just discharge 200 degree water. That doesn't really work anyway since our process cooling water has to stay at ~70 degrees.

We have some evaporation loss, but in the Phoenix climate it would be a lot more.

Right, I think the gp comment is referring to fabs being impacted by water shortages, not fabs causing water shortages.

Although if push comes to shove I guess all of this can be solved trivially by not growing ridiculous water intensive crops and not having golf courses and stuff in the middle of the desert.

Tucson resident here. I see acre after acre of alfalfa being grown, with most of that water then being exported to Saudi Arabia for cattle consumption. There is a geopolitical aspect to this, but I will gladly choose Arizona fabs over exporting cattle feed.

Tucson has been doing a very good job recharging our local aquifers and we also had a good monsoon. I was able to harvest 15,000 gallons of rainwater for irrigation on my land.

I view the construction of the fabs as critical infrastructure . Saudi cattle, not so much.

> Tucson resident here. I see acre after acre of alfalfa being grown, with most of that water then being exported to Saudi Arabia for cattle consumption.

(also Tucsonian) Whats really sad is we still have to search far and wide for decent quality hay to feed to our horses...

The fab investments also create a nexus of interesting jobs related to construction and fab operations. I really enjoy working in the latter sector.

That is not how that works. Las vegas is full of pools and fountains... INDOORS. they retain the moisture in their enclosed ecosystems. We have ACs that pull water out of the air, and we also know how to make buildings.

You can't just have AC when it comes to industrial cooling. You have to have chillers connected to cooling towers and those work on evaporative cooling. Feel free to debate all day on this though...I literally run the department in charge of it at a semiconductor manufacturing plant.

Here is proof, it's just some of the cooling towers at Intel's current site in Chandler: https://www.google.com/maps/@33.2451403,-111.8923691,135m/da...

So is the problem then that you can't push enough heat into the air, so you need to evaporate water? Sounds like a nice input to a desalination plant. Does this mean you agree that Arizona is a strange choice?

If they need so much water, does that mean you agree that

Absolutely it's a strange choice. From a power perspective it can make sense since solar works very well, but from a water perspective it's a very strange choice. Cheap land is also a very big factor. TSMC purchased over a thousand acres. Where else in the US can you find a thousand acres so near a populous area and so easily accessible off a major freeway like that?

Tax promises and Trump trying to sway the state for the 2020 election had more to do with the choice than anything I think.

the west coast also has a pretty low incidence of natural disasters. The east coast regularly gets bad hurricanes, the northeast and midwest get really bad snowstorms, plains states regularly get tornadoes, etc, but on the west coast if you pick a region that's not geologically active then there aren't really a ton of huge natural disasters that occur regularly.

of course I guess there's wildfires now too, but that's not really an Arizona thing either.

of course you're not wrong about companies often being lured by the particular states that are willing to offer them massive tax breaks, in some cases even to places like Texas that do get hurricanes on an occasional basis. And that doesn't always work out well in the end like with the Texas power outages that seem to be occurring more and more frequently, there are a LOT of fabs in Dallas/etc that are having to deal with widespread power outages multiple times a year.

(you're the expert here but it seems like the generator capacity usually isn't sufficient to continue normal operation of the fab, it's more to maintain containment/purity of the feedstock and you still lose wafers that were in-process at the time? that's the impression I've gotten at least)

No semiconductor site has enough backup generators to run continuously and you can only have so much diesel onsite to run those generators (a few hours or maybe a full day if you are lucky). They are there for basic safety systems, air handling and lighting. For example Samsung was down for many weeks after the February storm that caused power outages and fuel delivery issues. They literally ran out of diesel to run the generators and their cleanroom air stopped circulating. It took weeks to get things back to normal and my understanding is getting the air quality back was the hardest part.

https://www.zdnet.com/article/samsung-reopens-austin-chip-pl...

Your comments in this thread have been really interesting to read and I thank you for taking the time to respond to us all!

> the northeast and midwest get really bad snowstorms

I’m curious what the supposed issues around snowstorms are. In the Midwest at least you typically lose a day of “work” (not going into the office) a couple times per winter and that’s only because of school closures. The roads are rarely an issue for more than a couple hours unless it’s REALLY bad storm.

I don’t see why you couldn’t keep running the fabs through that.

Power bumps are a major issue and even with redundant power you get them. Even a bump that loses 15-20% of voltage on a single phase for 3-6 cycles (of 60Hz, so like 0.05-0.1s) is a huge impact to fab production tools.

I've seen impacts from cars hitting telephone poles 20 miles away that have been enough to impact the fab, let alone tree branches snapping from snow and ice.

I guess I don't know why this would be an issue - you just pick an area that uses buried power lines, and there are plenty of them. And ignoring that - if you're Intel building a fab, you setup shop near one of the major transmission lines that run well above the treeline. Or build your fab just down the street from the power plant...

St. Cloud, MN. Palo, IA. Brownville, NE. Burlington, KS. - nuclear plants.

I'm sure they all have plenty of cheap land to put fabs up on and if you take one of the norther states you'll cut your cooling bill to less than zero for several months out of the year without any evaporation.

Just having underground lines doesn't stop everything...and no one has underground lines for extremely high voltage.

re: West coast - one of the big wild fires got within five miles of Cupertino two years ago, local friends got evacuation notices before the fire was stopped, and all homes and I imagine even manufacturing buildings have to be built to withstand an earthquake all along the coast.

Everything is built for seismic safety, but that doesn't mean you won't have to scrap a ton of wafers or have any damage if there is an earthquake.

This is super interesting. Since you seem to have inside knowledge can you compare it to agriculture?

For example: the google machine tells me that cotton production in AZ uses somewhere in the neighborhood of 2.5-3.5 acre ft. water per acre. Would a large fab consume a comparable amount of water?

1 acre = 43.5k sq ft 1 acre foot = 325k gallons

In this case I can't compare it to agriculture. I don't think growing food in a desert is a good idea either.

The Intel site in Portland, OR uses ~2 Billion gallons of water per year. They claim they are bringing enough recycle capacity to save about half that.

https://www.oregonlive.com/silicon-forest/2017/08/intel_wate...

That’s not much. The average family of 4 in Arizona uses about 325,000 gallons a year. So a billion gallons is equivalent to 3,000 families, or a very small town.

Doesn't smell right, you're saying a family of 4 in Arizona uses ~900 gallons a day? That's 3500L or 3.5 tons of water every day.

I mean you can flush the toilet every minute for the whole day and only use about that much water, and that's assume the toilet fills up as fast (it doesn't).

That number seems really high, but the state website seems to be at least close to that number (146gal/day/person or ~600gal/day for a family of 4).

https://new.azwater.gov/conservation/public-resources

How to use a lot of water:

a) Water the lawn (in the summer)

b) Take a long shower

c) Drain and fill your pool

It probably includes yard irrigation.

Wow! I had no idea. Thanks for the thoughtful reply

Is it possible to do a closed cycle system? Or is that simply too expensive?

What exactly are you thinking? We have a closed loop system on our process cooling water but it is cooled via heat exchangers that use chilled water (in the 40-50 degree range). That water is chilled with extremely large chillers that use condenser water that is run through the cooling towers. It's as much of a closed system as you can get...eventually you need to remove heat and you can't just add energy to get it.

Wondering if it's reasonably possible to remove that last water step that is relying on open air evaporative cooling. Air cooled through big ass fans == no more water losses == profit?

Sounds like it would be multiples more expensive to get that to work though.

I don’t know about a fab, but the design principle I have heard from people making better use of residential and agricultural water is not trying to get it to a complete 100% closed loop, but to get multiple uses (yields) out of it as it flows through the system.

In a residential setting, it is typically a single-use where water purified to drinking water standards are used to flush toilets and water the lawn.

The idea here is to make better use of inputs and outputs. The folks in the Earthship projects just outside of Taos, NM build residences that cycle water 3-4 times before it goes out through a septic field, which feeds vegetation.

In none of those cycles, are the water re-used for the same purpose. You don’t try to purify grey water from bathing to reuse for bathing or washing, and instead, it is treated onsite through plants to reuse for things like flushing toilets. Water transports things, so what you look at is what is the waste being carried and how can that “waste” be used as an input for something else.

In the case of the fab, it sounds as if the major waste being carried is heat, via evaporative cooling. As I have mentioned in a different comment, that likely changes the microclimate near the fabs. So one area to look at are what kind of yields we can get with that changed microclimate? It depends upon the design of the cooling towers, and weather the water carrying the waste heat was still liquid when it got to the tower. Maybe there is another use for heated water, not specifically in the fab process itself, but something that could benefit humans and the ecology in a different way. For example, diverting some of that water to areas where zone 10 (tropical) plants are put into cultivation.

Taiwan went through a major drought due to going an entire year without a typhoon. TSMC is investing in recycling and water management.

https://www.taiwannews.com.tw/en/news/4292388

So I would assume Intel is also doing similar.

It looks like evaporative cooling is not the only thing Intel does. Recharging the aquifer at the Phoenix site is another: https://19january2017snapshot.epa.gov/www3/region9/waterinfr...

Depending upon how they do this and how much they are adding back into the aquifer, that can have a positive impact on the local ecology.

According to those slides, those were not the only strategies Intel used at the Phoenix site. It also added that there are already many water reductions strategies in use (which I agree with), so it doesn’t require groundbreaking research to develop.