Without getting too into the Ag weeds, every piece of American beef you have ever eaten is "Grass Fed". Beef cattle are, broadly speaking, born and raised on ranches where they eat rather copious amounts of grass in pastures. These ranches tend to be located in the plains states on land that is pretty much only good for grass, because if it was good for growing higher-value crops they would do that. If you would like an image, google "Nebraska Sandhills". That is the type of place where most beef cattle spend most of their lives.
When they go to the feedlots for finishing before slaughter is when they consume the most grain (corn) which is mostly grown on non-irrigated land (only 11% of US corn is irrigated). Producers try everything in their power to avoid feeding cattle until finishing because feeding cattle is much more expensive than letting them wander around a pasture. The vast majority of hay is grown on low-value land because if the land was better they would grow higher-value crops.
Isn't it strange that we (rightly) often regulate issues such as sediment, but at the same time currently have zero regulation of AI model development (or even real discussion of such regulation), when I think even the greatest AI optimist would acknowledge that the risks here are on a whole other level?
I will push back very slightly on your take on Chinatown. I recently rewatched the movie in the theater and the villain, Noah Cross, does something very different than diverting water from agricultural use. His plan was to create a fake drought in both the city and the rural areas in the northwest valley by dumping water into the ocean. Then once the farmers had lost so much money from the drought they were forced to sell their land, he used the phony identities of recently deceased people to buy a bunch of farmland on the cheap, floating below the radar. In addition to bribing politicians, he then used the 'drought' as a pretext to convince the city to build a reservoir on the land he just purchased at well below market rate.
The villainousness of his actions (besides the other, more clearly evil stuff he does) is more about defrauding both the farmers and the residents of the city. This is based on the actual actions of people like William Mulholland and Frederick Eaton, which were apparently quite similar. I think of the movie as more a criticism of the corruption and violence strangling the city at the time than the specific water-use issues at play. That plays into the other themes (and even the title) of the movie. What these people did was pretty clearly illegal, even if you believe (as I do) that it was the correct policy. I think the water itself is more of a 'based-on-a-true-story' backdrop than the main point of the film but I am completely open to disagreement, as silly as the whole thing is. Side note: I love old movies and I am more than happy to discuss film noir anytime.
I think it’s important to consider the effects of desalination on the ocean and its inhabitants. Pouring salt back into the ocean without considering the consequences seems short sighted. All technologies come with a down side. Let’s not perpetuate the belief that we should leave those problems to future generations. There’s enough evidence around to disprove that mindset.
Isn't it also the case that most water used by data centers is for cooling, so it is isn't really "consumed" in any sense: it's simply warmed and then sent back into the environment (the old refrain that you can rarely destroy water)?
I am more worried about energy use than water use. Data centers and such are only using about 1% of US energy but that is predicted to triple in a few years, and nobody can predict it.
And, yes, it this increase could be met with alternatives that would be nice, but a problem I see in this debate is that what we are seeing along with the rise in alternative energy is a greater rise in energy demand so the reality is that we continue to use fossil fuels now and the acceleration isn't slowing. The fact that a higher percent is coming from alternatives does not offset the fact that energy demand is increasing faster than alternatives are coming on line. https://ourworldindata.org/grapher/global-fossil-fuel-consumption
So, yes, wouldn't it be nice if the data centers could use alternatives, but that is true of everything but that doesn't mean they will.
Now, another totally useless energy waste is in generating crypto, crypto mining, a nonsensical and harmful pursuit that uses enormous computer power to do work that then becomes crypto. It now accounts for 1-2% of total energy use. They do need some work function to put a cost on new crypto, but they could as easily award crypto to how much plastic you can remove from the ocean, than on what they have now, solving math problems using enormous computer work, which is energy expensive
"You would need to increase American electricity output by 14 percent to desalinate enough water to meet our current irrigation needs. That’s not economically viable."
Not disagreeing, but...in 20-30 years, when we've electrified most other uses of energy, that 14% will be more like 5%, and also the cost of renewable electricity production will be even lower than today. At what point on this curve does it *become* viable?
"...increase American electricity output by 14 percent to desalinate enough water to meet our current irrigation needs."
Does this strike anyone else as a surprisingly small amount of electricity? I assume this omits electricity needed to pump desalinated water to places it's actually useful, but my main takeaway is that we could meet non-agricultural water requirements in metro areas near the ocean pretty easily with desalination.
Maybe our focus should be on:
-Not making the water rights situation worse in the West.
-Slowly chipping away at the most absurd water uses, for example by placing an export tax on alfalfa and having USDA/state ag extensions stop providing assistance for things like growing almonds in the desert.
-Making it very, very easy to build solar energy infrastructure and desalination plants.
Shouldn’t it be possible to fix the problem of “data centers use too much water” in the data centers themselves? I mean, “we’re going to desalinate some ocean water and pump it to a data center who will use it to remove heat and then dump it back into the ocean” is the most ass-backwards way I can possibly imagine to deal with heat issues and points to economic incentives being badly misaligned. Really data center operators should be incentivized/required to reuse the water, and air-to-water heat pumps exist with COPs of like 5+ that should be able to move the heat out of the water and into the atmosphere.
It appears that it's usually cheaper to cool with an open-loop system than with a closed-loop system.
My guess is that a lot of areas don't limit groundwater pumping because nobody has ever really pumped enough water to cause problems before. Data centers are now doing that.
Like you point out, the obvious answer is to just charge for pumping groundwater beyond some minimal amount for domestic usage.
not to be too pedantic, but the point about only 5% of beef being "grassfed" really doesn't inducate what you are claiming.
almost all us beef breeding stock is grazing on forage during the active growing season. the grassfed/grainfed distinction really only applies to the growth and finishing of the animal to be slaughtered which is a relatively small part of the total plant matter needed to produce a harvestable animal.
of course that also doesn't tell the whole story since a lot of beef is a byproduct of dairies, which mostly do feed crops year round...
at any rate, far more than 5% of beef is supported by grazing as opposed to planted and harvested crops.
I live in Tucson Arizona (essentially Arrakis) and there is a lot of drama about "Project Blue", a data center project that is being proposed.
I haven't done the most research on the project, but how much does building these data centers in drought ridden areas change the calculus on the "plenty of water for data centers". And while I agree there are plenty of ways to reduce water waste (for farming or whatever) those are politically unpopular.
I agree that water cost of using LLMs is probably overblown, but an analysis that only considers inference-time costs is missing a big slice of the pie. These models seem to have a very finite lifetime, so their amortized training costs are not negligible at all.
I'm probably just stupid and/or need a more physics-side explaining on water usage because from a very basic standpoint, water used in electricity generation isn't, like, destroyed. Even, as in most cases, it's heated to become steam, it will still, very quickly, become water again.
My understanding is that it's the where of the water. Let's say it gets evaporated - it'll come back as rain. But if you're getting that water from an aquifer, it won't all get back there immediately - some of it goes elsewhere. And some of that that goes elsewhere will become unusuable because it's salt or brackish water. The total amount of water's the same but the amount of usable water is down. Now, for a spoiled Southerner like myself this isn't a huge issue—we have water out the wazoo. But I'm sure for Westerners it could be more of an issue.
Totally! But then it just becomes a problem of moving water, or waiting for the timescale of replenishment. Both seem more tractable than the implication many seem to have that the water is gone/used up.
Also, I wonder if this problem is lessened due to climate change which is actually making rain more frequent.
Lastly, on the used fresh water becoming salt water you mentioned. I am no expert, but this seems like it cycles evenly. Otherwise we would have been seeing a long-run trend of the water on earth becoming salt water, on average. But I don't believe we do.
Would like to point out that the villainy in Chinatown is the rerouting of water away from farms in the San Fernando Valley and then buying up that arid farmland on the cheap for residential development that would be supported by a reservoir that was secretly being planned by the same people turning off the water.
Without getting too into the Ag weeds, every piece of American beef you have ever eaten is "Grass Fed". Beef cattle are, broadly speaking, born and raised on ranches where they eat rather copious amounts of grass in pastures. These ranches tend to be located in the plains states on land that is pretty much only good for grass, because if it was good for growing higher-value crops they would do that. If you would like an image, google "Nebraska Sandhills". That is the type of place where most beef cattle spend most of their lives.
When they go to the feedlots for finishing before slaughter is when they consume the most grain (corn) which is mostly grown on non-irrigated land (only 11% of US corn is irrigated). Producers try everything in their power to avoid feeding cattle until finishing because feeding cattle is much more expensive than letting them wander around a pasture. The vast majority of hay is grown on low-value land because if the land was better they would grow higher-value crops.
TL:DR American Beef is mostly born/raised on low-value, non-irrigated pastureland, they are mostly grain-finished with corn grown in non-irrigated fields. The vast majority of water used in cattle raising is rain falling into fields. Here's a fun map for where almost all of it happens. https://www.nass.usda.gov/Publications/Highlights/2024/Census22_HL_Cattle%20and%20Cattle%20on%20Feed_final.pdf
Isn't it strange that we (rightly) often regulate issues such as sediment, but at the same time currently have zero regulation of AI model development (or even real discussion of such regulation), when I think even the greatest AI optimist would acknowledge that the risks here are on a whole other level?
I will push back very slightly on your take on Chinatown. I recently rewatched the movie in the theater and the villain, Noah Cross, does something very different than diverting water from agricultural use. His plan was to create a fake drought in both the city and the rural areas in the northwest valley by dumping water into the ocean. Then once the farmers had lost so much money from the drought they were forced to sell their land, he used the phony identities of recently deceased people to buy a bunch of farmland on the cheap, floating below the radar. In addition to bribing politicians, he then used the 'drought' as a pretext to convince the city to build a reservoir on the land he just purchased at well below market rate.
The villainousness of his actions (besides the other, more clearly evil stuff he does) is more about defrauding both the farmers and the residents of the city. This is based on the actual actions of people like William Mulholland and Frederick Eaton, which were apparently quite similar. I think of the movie as more a criticism of the corruption and violence strangling the city at the time than the specific water-use issues at play. That plays into the other themes (and even the title) of the movie. What these people did was pretty clearly illegal, even if you believe (as I do) that it was the correct policy. I think the water itself is more of a 'based-on-a-true-story' backdrop than the main point of the film but I am completely open to disagreement, as silly as the whole thing is. Side note: I love old movies and I am more than happy to discuss film noir anytime.
I think it’s important to consider the effects of desalination on the ocean and its inhabitants. Pouring salt back into the ocean without considering the consequences seems short sighted. All technologies come with a down side. Let’s not perpetuate the belief that we should leave those problems to future generations. There’s enough evidence around to disprove that mindset.
The water use/AI issue is just another way for degrowthers to virtue signal. Like protesting multifamily developments or driving a Subaru.
Isn't it also the case that most water used by data centers is for cooling, so it is isn't really "consumed" in any sense: it's simply warmed and then sent back into the environment (the old refrain that you can rarely destroy water)?
It takes time and energy to make it clean enough to use and there's a finite capacity to make that.
I am more worried about energy use than water use. Data centers and such are only using about 1% of US energy but that is predicted to triple in a few years, and nobody can predict it.
And, yes, it this increase could be met with alternatives that would be nice, but a problem I see in this debate is that what we are seeing along with the rise in alternative energy is a greater rise in energy demand so the reality is that we continue to use fossil fuels now and the acceleration isn't slowing. The fact that a higher percent is coming from alternatives does not offset the fact that energy demand is increasing faster than alternatives are coming on line. https://ourworldindata.org/grapher/global-fossil-fuel-consumption
So, yes, wouldn't it be nice if the data centers could use alternatives, but that is true of everything but that doesn't mean they will.
Now, another totally useless energy waste is in generating crypto, crypto mining, a nonsensical and harmful pursuit that uses enormous computer power to do work that then becomes crypto. It now accounts for 1-2% of total energy use. They do need some work function to put a cost on new crypto, but they could as easily award crypto to how much plastic you can remove from the ocean, than on what they have now, solving math problems using enormous computer work, which is energy expensive
"You would need to increase American electricity output by 14 percent to desalinate enough water to meet our current irrigation needs. That’s not economically viable."
Not disagreeing, but...in 20-30 years, when we've electrified most other uses of energy, that 14% will be more like 5%, and also the cost of renewable electricity production will be even lower than today. At what point on this curve does it *become* viable?
"...increase American electricity output by 14 percent to desalinate enough water to meet our current irrigation needs."
Does this strike anyone else as a surprisingly small amount of electricity? I assume this omits electricity needed to pump desalinated water to places it's actually useful, but my main takeaway is that we could meet non-agricultural water requirements in metro areas near the ocean pretty easily with desalination.
Maybe our focus should be on:
-Not making the water rights situation worse in the West.
-Slowly chipping away at the most absurd water uses, for example by placing an export tax on alfalfa and having USDA/state ag extensions stop providing assistance for things like growing almonds in the desert.
-Making it very, very easy to build solar energy infrastructure and desalination plants.
Shouldn’t it be possible to fix the problem of “data centers use too much water” in the data centers themselves? I mean, “we’re going to desalinate some ocean water and pump it to a data center who will use it to remove heat and then dump it back into the ocean” is the most ass-backwards way I can possibly imagine to deal with heat issues and points to economic incentives being badly misaligned. Really data center operators should be incentivized/required to reuse the water, and air-to-water heat pumps exist with COPs of like 5+ that should be able to move the heat out of the water and into the atmosphere.
It appears that it's usually cheaper to cool with an open-loop system than with a closed-loop system.
My guess is that a lot of areas don't limit groundwater pumping because nobody has ever really pumped enough water to cause problems before. Data centers are now doing that.
Like you point out, the obvious answer is to just charge for pumping groundwater beyond some minimal amount for domestic usage.
not to be too pedantic, but the point about only 5% of beef being "grassfed" really doesn't inducate what you are claiming.
almost all us beef breeding stock is grazing on forage during the active growing season. the grassfed/grainfed distinction really only applies to the growth and finishing of the animal to be slaughtered which is a relatively small part of the total plant matter needed to produce a harvestable animal.
of course that also doesn't tell the whole story since a lot of beef is a byproduct of dairies, which mostly do feed crops year round...
at any rate, far more than 5% of beef is supported by grazing as opposed to planted and harvested crops.
I live in Tucson Arizona (essentially Arrakis) and there is a lot of drama about "Project Blue", a data center project that is being proposed.
I haven't done the most research on the project, but how much does building these data centers in drought ridden areas change the calculus on the "plenty of water for data centers". And while I agree there are plenty of ways to reduce water waste (for farming or whatever) those are politically unpopular.
I agree that water cost of using LLMs is probably overblown, but an analysis that only considers inference-time costs is missing a big slice of the pie. These models seem to have a very finite lifetime, so their amortized training costs are not negligible at all.
I'm probably just stupid and/or need a more physics-side explaining on water usage because from a very basic standpoint, water used in electricity generation isn't, like, destroyed. Even, as in most cases, it's heated to become steam, it will still, very quickly, become water again.
My understanding is that it's the where of the water. Let's say it gets evaporated - it'll come back as rain. But if you're getting that water from an aquifer, it won't all get back there immediately - some of it goes elsewhere. And some of that that goes elsewhere will become unusuable because it's salt or brackish water. The total amount of water's the same but the amount of usable water is down. Now, for a spoiled Southerner like myself this isn't a huge issue—we have water out the wazoo. But I'm sure for Westerners it could be more of an issue.
Totally! But then it just becomes a problem of moving water, or waiting for the timescale of replenishment. Both seem more tractable than the implication many seem to have that the water is gone/used up.
Also, I wonder if this problem is lessened due to climate change which is actually making rain more frequent.
Lastly, on the used fresh water becoming salt water you mentioned. I am no expert, but this seems like it cycles evenly. Otherwise we would have been seeing a long-run trend of the water on earth becoming salt water, on average. But I don't believe we do.
I was waiting and waiting for the issue of water rights to come up. Seems relevant when discussing water out west.
Would like to point out that the villainy in Chinatown is the rerouting of water away from farms in the San Fernando Valley and then buying up that arid farmland on the cheap for residential development that would be supported by a reservoir that was secretly being planned by the same people turning off the water.