Latest / Connor with Honor: AI, Real Estate and Real Life / The Shocking Water Crisis Lurking Behind AI Data Centers: And How It Could Dry Up Your Future (Unless We Fix It Now)

Transcript

The Shocking Water Crisis Lurking Behind AI Data Centers: And How It Could Dry Up Your Future (Unless We Fix It Now) · Connor with Honor: AI, Real Estate and Real Life

• 0:00
  SPEAKER_00: History tells us that when the people of France cried out, Queen, Queen, we are starving, we have no bread, Marie Antoinette supposedly replied, Let them eat cake.
• 0:10
  SPEAKER_00: Now flip that for today's world.
• 0:12
  SPEAKER_00: Imagine the people shouting, Mr.
• 0:14
  SPEAKER_00: President, Mr.
• 0:15
  SPEAKER_00: President, we have no water to drink.
• 0:17
  SPEAKER_00: And the reply, let them drink Diet Coke.
• 0:21
  SPEAKER_00: That's the kind of thinking we risk if we ignore how much water AI data centers actually consume.
• 0:27
  SPEAKER_00: So let's talk about that.
• 0:29
  SPEAKER_00: What is water usage effectiveness, WUE?
• 0:32
  SPEAKER_00: That's the main measurement that these data centers use regarding how much water that they're putting through the system in the cooling effort.
• 0:41
  SPEAKER_00: That's water usage effectiveness, WUE.
• 0:45
  SPEAKER_00: That means total gallons of water used for cooling divided by total kilowatt hours of energy used by the servers and IT equipment.
• 0:53
  SPEAKER_00: A lower W UE means less water used for the same amount of computing.
• 0:58
  SPEAKER_00: Many sitters have various systems.
• 1:00
  SPEAKER_00: Some are going to use a lot more than others, but I'll get into that in just a moment.
• 1:04
  SPEAKER_00: Most centters operate around 1.9 gallons per kilowatt hour.
• 1:08
  SPEAKER_00: It's a little confusing, but I'll kind of wrap the big numbers together.
• 1:11
  SPEAKER_00: So a one gigawatt data center, that's kind of the thing everybody's talking about, that equates to 1 million kilowatts.
• 1:18
  SPEAKER_00: 1 gigawatt?
• 1:19
  SPEAKER_00: 1 million kilowatts.
• 1:21
  SPEAKER_00: Okay.
• 1:21
  SPEAKER_00: Over 24 hours, that's 24 million kilowatt hours.
• 1:25
  SPEAKER_00: So at the W UE, that water expenditure number of 1.9 gallons per kilowatt hour, that's 12 million fifty thousand gallons per day.
• 1:35
  SPEAKER_00: So let's say 12 million gallons of water per day.
• 1:39
  SPEAKER_00: At a WUE of 0.7 gallons per kilowatt hour, that's a more efficient design of a data center.
• 1:45
  SPEAKER_00: That comes down to about 4.5 million gallons of water per day, 4.5 million gallons.
• 1:51
  SPEAKER_00: You got 12 million to around 4.5 million gallons of water per day, depending on how efficient the systems are.
• 1:58
  SPEAKER_00: So a single one gigawatt AI center drinks between 4 million and 12 million gallons of water per day.
• 2:04
  SPEAKER_00: So then my question was, well, well, is it is it is it is there a way to like recycle some of the water?
• 2:10
  SPEAKER_00: Does the water like evaporate?
• 2:11
  SPEAKER_00: Does it go somewhere?
• 2:12
  SPEAKER_00: Can it be trapped in like a cloud system and reproduce more rain, or is it just completely burned up and obliterated?
• 2:18
  SPEAKER_00: I'll talk about that.
• 2:19
  SPEAKER_00: But I also wanted to get into some different city comparisons.
• 2:22
  SPEAKER_00: So if you look at Los Angeles, they're at 426 million gallons per day for usage, a city like LA.
• 2:29
  SPEAKER_00: You have New York City, 1.1 billion gallons per day.
• 2:32
  SPEAKER_00: This is for people and stuff.
• 2:34
  SPEAKER_00: Phoenix, Arizona, 200 to 300 million gallons of water per day.
• 2:38
  SPEAKER_00: San Francisco, tens of millions.
• 2:40
  SPEAKER_00: So 12 million gallons a day is 2.8% of the Los Angeles water expenditure.
• 2:45
  SPEAKER_00: 1% of New York City, 4-6% of Phoenix, a large share of San Francisco.
• 2:49
  SPEAKER_00: Scaling to 10 gigawatts from one.
• 2:52
  SPEAKER_00: Now, 10, that's talked about.
• 2:53
  SPEAKER_00: Terawatt is as well, but 10 gigawatts, that's 120 million gallons of water per day at the high end.
• 3:02
  SPEAKER_00: One terawatt facility, that's 12 billion gallons of water per day at the high end.
• 3:07
  SPEAKER_00: At terawatt scale, we're talking about daily water use rivaling entire regions for water usage for people.
• 3:16
  SPEAKER_00: Now, here's where I wanted to talk about the cooling methods and water impact.
• 3:19
  SPEAKER_00: So I've heard of closed and open.
• 3:21
  SPEAKER_00: So the open systems, they use the cooling towers.
• 3:23
  SPEAKER_00: They spray water, let it evaporate like a swamp cooler.
• 3:27
  SPEAKER_00: They constantly refill, they really, really suck a lot of water.
• 3:30
  SPEAKER_00: Closed systems, those sealed loops, water recirculates in a loop, heat is moved to radiators or dry coolers, and much less water is used.
• 3:38
  SPEAKER_00: Then you have immersion cooling servers submerged in safe fluid, heat's carried away without evaporation, almost no water use in those types of systems.
• 3:48
  SPEAKER_00: The least expensive, of course, less water efficient, more expensive, more water efficient.
• 3:54
  SPEAKER_00: There's a cost thing.
• 3:55
  SPEAKER_00: So then I said, well, what about salt water?
• 3:57
  SPEAKER_00: We have quite a bit of that on the planet, not like I want them to use it all because we get rain from that, but it's abundant, doesn't drain fresh water supplies, but you have the corrosion of equipment, mineral buildup being on the bad end of it, biological fouling, costly filtration, strict environmental discharge rules.
• 4:13
  SPEAKER_00: So some coastal centers with the AI development of these data centers, they're using these hybrid salt systems, but it's expensive and technically difficult to maintain.
• 4:24
  SPEAKER_00: In the future, room temperature superconductors, RTSs, room temperature superconductors, I've heard this talked about before.
• 4:31
  SPEAKER_00: If scientists develop them, these room temperature superconductors, it's fun to say, materials that carry electricity with zero resistance at everyday conditions.
• 4:40
  SPEAKER_00: You have no wasted energy as heat, much less cooling is required.
• 4:43
  SPEAKER_00: Water demand drops dramatically.
• 4:45
  SPEAKER_00: But it's speculative for now.
• 4:46
  SPEAKER_00: If RTS becomes real, it could transform the water footprint of data centers.
• 4:51
  SPEAKER_00: Then give us the only other issue that we seem to be butting up against, which is how energy is going to be supplied, because these are very energy hungry systems.
• 5:00
  SPEAKER_00: That's another video.
• 5:01
  SPEAKER_00: How many one gigawatt data centers exist or planned?
• 5:03
  SPEAKER_00: Well, here's what we know global capacity in 2024.
• 5:07
  SPEAKER_00: That's last year.
• 5:08
  SPEAKER_00: We're now closing out September 2025 to give you an idea.
• 5:11
  SPEAKER_00: This stuff changes so dramatically.
• 5:12
  SPEAKER_00: Got almost to a show every day.
• 5:14
  SPEAKER_00: You have 122 gigawatts totally across all data centers as in 2024.
• 5:20
  SPEAKER_00: Some regions, Northern Virginia, has already exceeded one gigawatt in total capacity.
• 5:25
  SPEAKER_00: A few individual sites have been announced above the one gigawatt rating, including some of Oracle's projects.
• 5:31
  SPEAKER_00: Multi-gigawatt campuses up to 5 gigawatt are being planned now.
• 5:35
  SPEAKER_00: And in 2025, about 10 gigawatt of new capacity is expected to break ground worldwide.
• 5:41
  SPEAKER_00: So that's this year.
• 5:42
  SPEAKER_00: By the end of this year, we should have that much.
• 5:45
  SPEAKER_00: By 2030, demand could reach 171 to 219 gigawatts of new energy.
• 5:52
  SPEAKER_00: And even with that, the US alone may still face a 15 gigawatt shortfall.
• 5:58
  SPEAKER_00: So a single one gigawatt data center is still rare, but they're real and more are coming.
• 6:05
  SPEAKER_00: So what do those amounts mean in water?
• 6:07
  SPEAKER_00: So at 12 million gallons a day, that's on the upper end of water expenditure, not a super efficient system.
• 6:14
  SPEAKER_00: That's 12 million gallons of water a day for each one gigawatt data center.
• 6:19
  SPEAKER_00: 10 centers that, of course, do the math, 120 million gallons of water per day.
• 6:24
  SPEAKER_00: That's about 28% of Los Angeles' daily water use.
• 6:27
  SPEAKER_00: A hundred of these centers, 1.2 billion gallons a day.
• 6:30
  SPEAKER_00: That's more than all of New York City's daily water requirement.
• 6:34
  SPEAKER_00: Every new gigawatt class center cuts into the same water supplies that household farms and rivers depend on.
• 6:41
  SPEAKER_00: Now I talked about rainfall.
• 6:42
  SPEAKER_00: I looked about into this and you know in essence it really doesn't give good answers and good comparison.
• 6:50
  SPEAKER_00: But if you have every year over land 103 quadrillion gallons of water falls as rain, and that's around 360 billion gallons per day of rain.
• 7:02
  SPEAKER_00: In the U.S., the average is about 30 inches per year, which creates roughly to 4,000 billion gallons per day nationwide or 4 trillion.
• 7:11
  SPEAKER_00: Yes, the water cycle is huge, but most of that rain falls in oceans, deserts, or the wrong places.
• 7:16
  SPEAKER_00: Data centers draw from drinking water systems and aquifers, not the random rain that falls on empty ground.
• 7:22
  SPEAKER_00: But of course, where do we get water?
• 7:23
  SPEAKER_00: Well, it comes from rain.
• 7:24
  SPEAKER_00: But anyway, it's I think it's more important to look at what they're actually taking because I think that's really going to be an issue, and we can outrun that rain at some point.
• 7:34
  SPEAKER_00: So we have to be wary.
• 7:35
  SPEAKER_00: Why water rarely gets attention, but must everybody talk about electricity, carbon emissions, chip supply, and something else about chip supply, if they do come in with this room temperature superconductor, these RTSs, if that is developed and becomes a thing, becomes a thing.
• 7:51
  SPEAKER_00: The question I would have, and I've tried to research this, how much of the current infrastructure has to be swapped out to allow this to work?
• 7:58
  SPEAKER_00: So water usage is cut to near zero by implementing these room temperature superconductors?
• 8:04
  SPEAKER_00: Is it the entire system?
• 8:06
  SPEAKER_00: Or are they going to be able to, let's say, keep these particular chipsets and these these uh RTSs operate somewhere else and help something else?
• 8:14
  SPEAKER_00: Or does the entire original system have to be scrapped that right now, of course, we don't have a new chips, enough chips to be able to put into these places anyway?
• 8:22
  SPEAKER_00: There's like backward lists, and people are paying premium to get, I mean, millions and millions and millions of chips.
• 8:29
  SPEAKER_00: So it it's it's like an episode of Star Wars.
• 8:32
  SPEAKER_00: I mean, they're just manufacturing, manufacturing, manufacturing.
• 8:35
  SPEAKER_00: What if somebody came up with this room temperature superconductor?
• 8:39
  SPEAKER_00: What does that look like?
• 8:40
  SPEAKER_00: What does the overlay look like on the current uh data facilities?
• 8:46
  SPEAKER_00: And are there going to be some kind of governmental regulations put in place?
• 8:49
  SPEAKER_00: Because right now I know California, I think there's an assembly bill that talks about this water usage having to be revealed when these data centers are built, at least in the state of California.
• 8:59
  SPEAKER_00: I think that's coming up.
• 9:00
  SPEAKER_00: Then what's going to make people want to build those in California when they can go somewhere else?
• 9:05
  SPEAKER_00: Some of the thing, another thing also is if you look at the land around mountains with lakes and streams and these sorts of things, I believe all that land is starting to get picked up if it's not gone already here at the end of September 2025, because they know that that water is going to be very necessary, at least with present technology, to keep these AI systems moving along in the way that they want.
• 9:27
  SPEAKER_00: At some point, hopefully the water part of the equation will get solved and people won't get thirsty and die of thirst because you can go a lot a lot longer without food than you can water.
• 9:37
  SPEAKER_00: However, then on the other side of that, how does that look when we have that issue solved, maybe with some kind of room temperature superconductors or putting these places maybe where they don't have that kind of a footprint that requires that much water, maybe in a cooler area, but then you have ice meltage and there's all of another issue there.
• 9:56
  SPEAKER_00: But the power, where does the power come from?
• 9:58
  SPEAKER_00: And people are looking right now at fossil fuel being the easiest supplier, at least the most convenient currently.
• 10:04
  SPEAKER_00: But the nuclear is a thing, and I know that they've reduced a lot of regulation on that, but that takes time apparently to build those out.
• 10:11
  SPEAKER_00: I mean, look at what Musk has done as far as time frame.
• 10:14
  SPEAKER_00: He really shrinks those scales, doing a lot of good work very, very quickly.
• 10:18
  SPEAKER_00: Is the government or other other uh builders of the technology going to be able to do the same?
• 10:23
  SPEAKER_00: And then the other question pulling back just a little bit, because that's great, right?
• 10:26
  SPEAKER_00: Don't you never want to not bet on Elon.
• 10:29
  SPEAKER_00: It's a bad idea, so they say.
• 10:30
  SPEAKER_00: But then should we be?
• 10:32
  SPEAKER_00: Should we be building so fast?
• 10:34
  SPEAKER_00: Are we missing something?
• 10:35
  SPEAKER_00: Are we leaving something out?
• 10:37
  SPEAKER_00: The water usage doesn't really get talked about.
• 10:39
  SPEAKER_00: They're all talking about the other things I mentioned, the electricity, carbon emissions, chip supply, and they're all very, very important.
• 10:46
  SPEAKER_00: You know, where we have to get the power, the people that need air conditioning, are they gonna get it?
• 10:51
  SPEAKER_00: The elderly, are they gonna get it?
• 10:52
  SPEAKER_00: What's gonna happen?
• 10:53
  SPEAKER_00: Are they gonna say, well, today we're gonna use the data center, give it the power, so you're gonna have to go without power for six hours, but it's only from six at night to 12 at night.
• 11:01
  SPEAKER_00: I mean, who uses power at those times, sort of thing?
• 11:04
  SPEAKER_00: And then almost nobody talks about the water.
• 11:06
  SPEAKER_00: Most operators don't publish their W UE or how much water those data centers are actually expected to use.
• 11:13
  SPEAKER_00: In drought zone, a one gigawatt data center could equate to a small city's water use.
• 11:18
  SPEAKER_00: The silence's not going to last.
• 11:20
  SPEAKER_00: Communities, I think, will push back.
• 11:22
  SPEAKER_00: You know, if we want AI without draining the aquifers and rivers, we're gonna have to require water usage effectiveness, reporting that WUE would need to mandate water-efficient cooling, more closed loops, immersion type system, or hybrid systems, use non-potable or recycled water, invest in new material and superconductors.
• 11:40
  SPEAKER_00: That's a big one.
• 11:41
  SPEAKER_00: I think that problem's probably gonna be solved more sooner than later.
• 11:44
  SPEAKER_00: But what's the cost?
• 11:45
  SPEAKER_00: You know, you you cure a disease that has a lot of money attached to it in the way that it keeps it, it keeps it from overtaking a person.
• 11:53
  SPEAKER_00: And, you know, small fixes, maybe in some cases it does cure in other cases.
• 11:58
  SPEAKER_00: You know, it might not help at all, but it is in place.
• 12:00
  SPEAKER_00: And if you find a solution or cure that's relatively inexpensive, it can be mass-produced for that certain disease.
• 12:07
  SPEAKER_00: Is it going to be released?
• 12:08
  SPEAKER_00: You know, in my generation, you always hear those conspiracies.
• 12:11
  SPEAKER_00: The newer people out there maybe have never heard something like that.
• 12:14
  SPEAKER_00: But in this case of these data centers, you have these massive, you know, at some point, these massive data centers, multi-gigawatt facilities, and you have them, and then you have them building a new uh room temperature conductor that's able to be utilized.
• 12:32
  SPEAKER_00: What does that change out look like?
• 12:34
  SPEAKER_00: Would that technology ever be released, or would it just kind of be patented and hid or whatever it may be?
• 12:40
  SPEAKER_00: Lots of questions.
• 12:41
  SPEAKER_00: Please answer in the comments.
• 12:43
  SPEAKER_00: I'd love to hear what you think on this.
• 12:44
  SPEAKER_00: That's a lot to think about.
• 12:47
  SPEAKER_00: But you know, AI is powerful.
• 12:49
  SPEAKER_00: Water's infinite.
• 12:50
  SPEAKER_00: Every one gigawatt data center adds millions of gallons to daily demand.
• 12:54
  SPEAKER_00: If we ignore that, tomorrow's AI boom could become tomorrow's water crisis.
• 12:59
  SPEAKER_00: So definitely let's scale smart.
• 13:01
  SPEAKER_00: Let me know what you think.
• 13:01
  SPEAKER_00: I'm Santa Clarita Artificial Intelligence.
• 13:04
  SPEAKER_00: I'm Connor with Honor, not just a clever name.
• 13:06
  SPEAKER_00: And I look forward to hearing your comments and interacting with you on the different platforms out there.
• 13:11
  SPEAKER_00: Be well.
• 13:12
  SPEAKER_00: We'll see you in the next one.
• 13:12
  SPEAKER_00: Thanks for watching.