Big tech in water
The AI boom and data centers.
Hello, and welcome to Western Water Notes.
To get all my posts in your inbox, click the button to subscribe below. This newsletter is free, but if you find my work valuable and want to support it, please consider the monthly or yearly subscription plans. You can also support my independent writing by sharing these posts with friends or on social media. As always, drop me a line with feedback or suggestions.
In August, Google pledged to spend $400 million this year on data centers near Reno and Las Vegas. The announcement, the company told the Reno Gazette Journal, would bring its investment in Nevada to $2.2 billion since 2019. Data centers today are the key nodes in our economy. Most of us interact with them, unknowingly and indirectly — almost every day. And big technology companies are investing billions in them.
It’s one reason tech spending is expected to reach ~$5 trillion this year.
But what caught my eye in the Gazette Journal story had less to do with the size of the investment and more to do with another factor: water. Nevada is the driest state in the nation, and data centers demand water for cooling. Google is hardly unaware of this. In fact, the tech company is getting into water in many of the places that it operates.
This detail stood out:
As part of the Reno deal, Google will invest $500,000 in tree thinning and vegetation management in the Tahoe National Forest to “retain water in the region,” the Gazette Journal article says (though there aren’t many more details about how this water would be accounted for). If I’m reading it right, Google will invest in the project in the hopes of making the system more efficient and offset its withdrawals elsewhere in the basin?
Its funding pledges don’t end there.
On the other end of the state, Google has spent $4.7 million (as of 2023) on several Colorado River Basin projects in an attempt to offset its water footprint at a data center in Southern Nevada (though again, it would be helpful to see more accounting of water saved vs. used). In late August, the Walton Family Foundation touted water efficiency projects on the Verde River in Arizona. Among the companies funding the projects were some of the biggest names in tech and AI: Google, Meta and Microsoft.
This post is not meant as a rant, really. In keeping with the goal of this newsletter, it’s simply meant to make the invisible more visible. That is, the role of big tech in water.
If data centers serve as economic junctions in a world of AI, the foundations are built as much on land as on water and power. The power side of this has been well-reported on and discussed. Microsoft is restarting Three Mile Island, as consultants speculate data centers could make up 12% of energy demand by 2030, as hard as it is to believe.
But if the water part has not received as much attention, it should.
Exactly how much water are we talking about?
Consider this in The Washington Post. Relying on UC Riverside research that looked at ChatGPT’s GPT-4 language model, reporters found the average 100-word email used a little more than 1 water bottle. That might not sound like a lot at first. But when AI is scaled, it is hardly an insignificant volume. If 10% of the U.S. workforce used GPT-4 to write an email once a week for a year, it would amount to all the water consumed by households in the state of Rhode Island for 1.5 days. And this is just to write an email.
As the article notes, the amount of water consumed varies from place to place. But it’s easy to envision a world (not far away) where AI could require more and more water.
Many (maybe even most?) technology companies have made pledges to invest in more efficient data center cooling. Google, Meta and Microsoft have all said they plan to be “water positive” by 2030. The term suggests they will replenish more water than they use for their operations. But as with all environmental credit, offset and mitigation programs, the pledge is only as good as the numbers. Validation, accountability and transparency are key. How do you measure replenishment? What does and does not count? This feels especially hard with water, as supplies can vary so much each year.
It feels especially hard when tech companies have not always been forthcoming.
In some cases, reporters and watchdogs have had to push for data. A city in northern Oregon sued The Oregonian after a reporter sought the release of Google’s water use data under public record laws. When the city of The Dalles settled its suit with the paper, the water data it released showed Google’s water use had tripled in five years.
Asking about the water
Over the past few weeks, I’ve heard various echoes of the same refrain: Those who do not learn how to use generative AI will be left behind. I have a lot of thoughts on that (for another post on another day) but I see where good-faith boosters are coming from.
The many applications data centers support are weaved into our lives. They support medicine and science. They support education and art. They are ubiquitous. After all, the platform I wrote about in my last post (OpenET) is only made possible by Google.
But that is the point. As with so much in our modern world of distant supply chains, we interact on a daily basis with data centers, yet we don’t know it. Our relationship is about to deepen dramatically with AI — and there are real-world consequences about how we consume water and power. It is worth thinking about that now, ahead of time.
And policymakers, reporters and the public should ask: What about the water?
What I’ve been reading:
People — and salmon — return to restored Klamath (Arizona Republic)
As California farms use less water, worries grow over Salton Sea (L.A. Times)
A receding Lake Powell brings Colorado River rapids in Utah back to life (NPR)
Can New Mexico’s ancient water system survive climate change? (Undark)
Feds won’t let anticipated sewage spill hit the Tijuana River (Voice of San Diego)
Federal solar plan at odds with decadeslong monument fight (Review-Journal)
More lithium projects approved in Utah’s redrock country (Salt Lake Tribune)
The 2024 state of the climate report: Perilous times on planet Earth (BioScience)
And a bonus science communication article: What’s happening in America today is something darker than a misinformation crisis (The Atlantic)