Comments on the NYT's most recent coverage of data centers and water
It's that time again folks, here's my reactions to the NYT's most recent piece on data centers and water.
I'm very disappointed.
First, the title and subtitle.
Microsoft data centers are not currently using that much water. The article notes that all Microsoft data centers used 10.4 billion liters in 2024. My hometown of 17,000 people uses about 2 billion liters of water each year. So all Microsoft data centers around the whole world will jump from the equivalent to 5 of my towns to 10 of my towns by 2030. Not nothing, but this is the big water catastrophe the whole article is about.
It's true that data centers are being built in places that face shortages. There's plenty of industries and farming that use water in places that face shortages. Like I say a lot, this doesn't give us much useful info on its own, but it's a subtitle so I guess we can't expect much info to be crammed in there.
Onto the article itself
Right away, I really hate this intro framing.
This reads like by going along with the AI boom, Microsoft is somehow taking away something essential to life. In reality, I don't expect them to harm anyone's access to water and instead to just participate in the normal cycle. Contrasting this quote with the "Then came the AI boom" line after paints such a silly picture.
Next we get 4 paragraphs full of contextless large numbers, my arch nemesis in environmental reporting. In an ideal world these would basically never make it past an editor's desk. They could at least give you a percentage of each region's water the numbers refer to. I am down on my knees: editors, please please please do not let contextless large numbers into the final edit of your piece.
If you actually look into how much water each region's using, the big numbers look a little smaller. If you include the groundwater Jakarta uses, the data centers will add about 0.03% to the city's annual water demand. The Phoenix number will leave data centers there using about 3% of the water used by the area's golf courses. Do you think readers will leave this piece with a reasonable sense of the proportions here?
Instead of proportions, the reader is greeted with some ominous pictures of each region.
The article then goes on to say that the AI boom's been caused by chatbots, and builds up to yet another big drop of a contextless large number.
I'm at least happy to see them citing the LBNL, the GOATs of estimating data center water use (for now). I do also want to give them credit here for sharing the LBNL's guess at direct water consumption. Sometimes articles blur the line between direct (onsite) and indirect (offsite in power plants) water use, and they're not doing that here. Luckily, they do contextualize this number a few paragraphs down:
Hey! Lots of good stuff here! A direct comparison to another industry! How much water this is using as a percentage of the US overall! Very happy to see this. I think the average reader is often left not knowing things like the fact that Toyota is using this much water to manufacture cars. I wish all articles had comparisons like this.
My question though is why leave this toward the end of the article? Why not just add the context where those big numbers are dropped?
The very next line unfortunately I think adds some confusion.
Okay, so the reader is left without any context for how big of a problem this is, because their one point of comparison has been invalidated as meaningless, and the article gives no hint of how big of a problem this actually is on the local level. The choice to call municipal water "drinking water" that's "evaporated away" I think also might leave readers not understanding that lots of other industries also use potable municipal water. And again, so far I have found zero areas where data centers have actually harmed water access, but I'm not sure readers are being left with a sense that that's the case.
Before moving on, I want to circle back on a quick note that happens between the last sections I talked about:
This is basically an argument that water-stressed regions shouldn't have industry. I don't think people would object if a car factory were opened near Jakarta (or near Phoenix). There's this very weird common implication that data center companies should choose not to build where lots of other industry is. In an ideal world, data centers would propose sites where it made economic sense to build, and local governments would price and regulate water well so that if it were actually bad to build a data center there, the water price would signal that and the company would be forced to build elsewhere. I don't think it's especially ominous that new industry is being built near Jakarta. If people want to live in a water-stressed region, it seems like it's okay that there's some taxable industry around them as well to sustain their communities.
The article then transitions to notes from an interview with Microsoft's former director of water strategy. This caught my eye.
Obviously we should try to conserve water and protect unstable water sources, but I think a lot of people underrate how cheap water already is in America. We have way less expensive water than almost any other wealthy country, to the point that companies using billions of liters per year don't even have to consider it as a priority. This quote I think pushes against the piece's broader theme that water access is threatened.
Later on in the same section:
Sorry, doesn't this mean there aren't that many urgent water issues in the places where Microsoft is building? It's not that this implies there aren't any issues at all, but this also subtly cuts against the thesis of the piece.
This next section made my goodwill toward the article drop quite a bit:
This is an incredibly common move in writing about this topic: report that data centers are built in water stressed regions, mention all the problems this water stress causes, but give your readers zero idea of how much the data centers are actually contributing to said water stress. This is a sloppy move. Huge numbers of industries that use water operate in water stressed regions. Arizona has golf courses and alfalfa farms and water parks. If people want to live in a water stressed region, they need industry and commercial buildings to tax and work at, and those use water! This whole section tells the reader nothing about how much of a problem these data centers are creating. It only ominously notes that it's a lot of water by the standards of the company's other data centers. Basically useless info. All Arizona data centers together use about 0.1% of the water used by Maricopa County alone, but provide way more tax revenue per gallon than the golf courses there, which use way more water in total. The reader has no way to compare data centers to other industries in these regions. They're just left with this ominous implication that the data centers shouldn't be there.
After some more ominous pictures, we get this:
That is the water demand of my hometown of 17,000 people. Again, not nothing, but Pune seems to use more water than this every single day. This is 0.3% of Pune's water demand.
The article then goes on to talk about electricity for a while, before closing with this line:
Again, leaves the reader without comparisons to make. If I were trying to leave the reader maximally informed about how much water data centers use and how they compare to other things, this is not the article I would have written.