What are data centers? Why are people so upset about them? Where does cloud data live on the internet? What are the consequences of the ongoing revolution in so-called “artificial intelligence,” or AI? And why are governments and corporations investing in AI so heavily, especially when the tech seems financially unviable?
In this episode, we hear answers to these questions from CELDF’s Co-Director of Community Resistance and Resilience and our staff attorney Will Falk.
Will is a lawyer and grassroots community organizer who has been working for decades to defend communities and the natural world. He has been part of blocking oil pipelines at the Unist’ot’en camp on Wet’suwet’en First Nation unceded territory in what is today known as British Columbia, took part in the movement to defend Mauna Kea from the construction of the 30-meter telescope (TMT) on the sacred Mauna, helped launch the first major lawsuit in US history in which an ecosystem sued to protect its own rights (Colorado River vs. The State of Colorado) and is the co-founder of Protect Thacker Pass, where he helped organize a direct action campaign that went on for three years and represented multiple Native American tribes in filing Federal Court lawsuits against a mining company.
Will is also an accomplished poet who recently finished his MFA in poetry, and is the author of two books. You can learn more about Will, and contact him, via his website, willfalk.org. He’s also active on Facebook.
Excerpt 1: The Hidden Water Cost of AI
“Every AI Query Uses a Bottle of Water”
Artificial intelligence and data centers can be really water-intensive. And of course, this varies based on scale. One of the primary uses for water for data centers is to cool them down. Whenever you have machines running on electricity, machines primarily made of metal, they’re gonna heat up and they can malfunction if they get too hot. So you have to have constant water to cool these machines down so that they can remain operating.
Here’s what the numbers show: a study from the University of Georgia found that a single artificial intelligence query is estimated to use 16 ounces of fresh water. ChatGPT alone estimates that it gets more than 10 million queries per day. So 10 million people typing in Google questions—that’s 10 million times 16 ounces of water being used just on query functions. That’s a lot of water.
To put that in perspective, scrolling through social media for one hour could use 15 ounces of fresh water. And a one-hour video conference call on Zoom uses approximately 60 ounces of water.
Right now, the world’s largest data center hub is in Loudoun County, Virginia. It used around one billion gallons of drinkable water in 2023, or about 2.7 million gallons per day.
One important thing to note: a lot of these numbers are estimates because data center corporations are not transparent about how much water they’re using. A lot of laws are not forcing them to be transparent. These data centers often hide the amount of water they’re using by classifying it as proprietary information, like business secrets. Local governments and others responsible for permitting these processes are signing non-disclosure agreements so they’re not telling their constituents how much water is being used.
So the numbers could be higher. Why else would corporations want to hide the water usage if it wasn’t because the numbers are huge? Water is life, and they are using the source of life to cool down their very hot machines.”
Excerpt 2: The Myth of “The Cloud”
“There Is No Cloud—Only Machines”
“What is artificial intelligence? On a very basic level, it’s when human experts program computers, program machines, to sift through large amounts of data and apply logical binaries. This is where you get the ones and zeros of digital action. You train a machine to apply these logical statements: if this happens, then this should follow.
Because these machines can analyze huge amounts of data much faster than a human mind can absorb that information, they can start to create an internal logic based off of consuming all of that information.
But the key to understanding this is recognizing that it’s not something that just happens up in the cloud. It’s not like just a radio wave. It’s not something where all this information is simply stored out in the sky somewhere. That’s not how it works.
Just like a human brain has to have a physical place to store all of our experiences and the information that we take in, the same thing is true for artificial intelligence. It needs these physical locations. It needs these physical machines to be able to process the information that informs artificial intelligence. And it needs sites where this information is physically stored.
That’s why artificial intelligence requires data centers. One reason this is becoming such an issue today is that artificial intelligence has been developed to a point where it needs so much more data to produce the kinds of things that we need. And it needs so much computational energy to come up with those data sets—the stuff that the machine spits out after you pump information into it.
Whenever we hear about new developments in artificial intelligence, whenever we hear artificial intelligence getting more powerful, what we also need to see is the machines getting bigger. That’s literally what happens. The data center spaces have to get bigger and bigger. The infrastructure that supports those data centers has to get bigger and bigger, more energy-consumptive, with more extraction necessary to create that infrastructure.
And at the same time, we have to see the expenditure of fuel, the consumption of fossil fuels in order to support those very energy-intensive machines.”
Excerpt 3: Organizing Against Data Centers
"“AI Infrastructure Can Be Stopped on the Ground”
“The good news is that because these processes require a large and massive infrastructure to operate—infrastructure that has to be kept running smoothly in order for the system to operate—that means we can directly challenge those projects. We can directly resist these projects on the ground. We can in fact dismantle those projects on the ground. They can’t guard all of the AI and data center infrastructure all the time.
A lot of this stuff is much more fragile than you would think. We’ve gotten used to this notion that fossil fuels are always there, they’re always gonna be there, they’re never gonna run out. Of course, we know if we step back and think about it, fossil fuels will run out. There’s only a finite amount on Earth. When you’re using more fossil fuels every year than the year before, eventually you’re going to take a slice of the pie and there’s no more pie there. If we continue on this trajectory, basic math says we will run out of fossil fuels. Artificial intelligence and data centers won’t work when that happens.
The reason I say this is fragile is that when you find a choke point, you can threaten the global economy and force it to slow down. This is also true with the materials used for these machines. Data centers and the machines that run artificial intelligence use very specific materials, and there’s not an infinite amount of those materials.
If we can interfere with those supply chains, if we can interfere with the extraction, if we can interfere with the transportation of energy and minerals, then we can slow down or even make it so that data centers were virtually economically impossible to produce.
The question of what can we do about it comes down to this: start to organize on the ground to oppose these projects where they’re happening. Think bigger about infrastructure and systems and supply chains and how we can interfere with those things. Make it harder for them to manufacture and get this stuff up and running. Because these processes exist out in the real physical world, out in communities on the ground, in people’s actual backyards—they can be stopped.”
Links and Resources
* “terminate with extreme prejudice” — Arund…