Adopt AI Today, Die of Thirst Tomorrow
Big Tech has been draining away the planet’s fresh water at an unsustainable pace. AI is making it even worse.
Earlier this year, Geoffrey Hinton, the “godfather” of artificial intelligence, quit his job at Google to do a press tour on the dangers of AI. The dangers he listed, of course, involve a sci-fi scenario where an autonomous, sentient AI decides to nuke our planet from orbit. Or, perhaps, simply flood our society with (even more) disinformation that will (further) impact our political elections.
Several brave women, like AI researchers Timnit Gebru and Meredith Whittaker, contend that the danger of AI is far more immediate—such as perpetuating racism or causing people to relinquish vital decision-making capabilities.
But the situation is even worse than that. The servers that sustain these AI models are rapidly consuming our planet’s supply of fresh water.
Big Tech’s Water Problem
The servers that house and run the world’s internet, email, social media platforms and cloud storage exert tremendous computing power. That power generates a great deal of heat. If you ever thought that your smart phone or laptop get too hot to handle at times, imagine the units running inside a data center or server farm. A single server rack (housing about 20 servers) can generate 12 kilowatts (kW) of heat—or 40,800 BTUs: the same as a residential gas furnace.
A human body generates about 400 BTUs.
So you’ve got these massive data centers housing thousands of these racks (possibly millions in Google’s case), and each rack is producing the BTU equivalent of a gas furnace. Unmitigated, that level of heat would melt the hardware.
So what do you do? Install some really hefty air conditioners—along with a liquid cooling system to suck away the heat.
The most economical liquid for this job? Fresh water.
How much water? Try 5.6 billion gallons used by Google and 1.7 billion gallons used by Microsoft. In 2022. Alone.
Granted, that’s a tiny percentage of the fresh water the U.S. uses daily. But that percentage is quickly rising.
Why Fresh Water?
Potable water is least corrosive to the equipment that makes up these cooling systems. While Big Tech companies could theoretically use specialized oil for the job, it’s more expensive and they’d have to repair and replace various parts of the system more frequently.
Why not simply recycle the water? Well, they do. The problem is, roughly half of it evaporates during the cooling process.
The water cycle we learned about in grade school illustrates the issue at play. The companies take fresh water from underground aquifers and other sources. A great deal of it evaporates from the high heat and escapes into the atmosphere, where it’s condensed and falls back to the earth’s surface as rain. Some of that rain falls into the ocean. Some of it is contaminated by chemical runoff. Some of it is reevaporated by warm surface temperatures (which are increasing due to climate change). And a little bit filters back through the soil to the aquifer.
As you might suspect, the rate of pumping is far exceeding the rate of replenishment. Google predicts that global demand for fresh water will outpace supply by 40% in 2030.
That’s only seven years away.
What is Big Tech doing about it?
You’d think this alarming water crisis would have Big Tech leaders scrambling for solutions. Instead, they’re continuing to develop and release various AI models, which is doubling server computing power every 100 days and on track to increase exponentially.
These companies, of course, have sustainability goals outlined in their annual reports. Google’s, for example, is to replenish 120% of its water consumption by 2030. Its current replenishment rate stands at 6%. Ignoring the fact that contributing more water than one uses is literally impossible (What are they going to do? Create water?), that’s a vast improvement to make in only seven years.
As the peasants say, the math ain’t mathing.
Exacerbating a Genuine Crisis
Per Big Tech’s modus operandi, they’re currently focusing on how they can make their processing and cooling systems more efficient, so less water per kW is needed. They also claim to be funding projects with NGOs to improve the water supply.
But with AI driving up their computing power, it’s difficult to see how these innovations could be substantive enough—especially when you consider that nearly half of U.S. aquifers are already approaching critically low levels. The situation is so dire that Phoenix, Arizona, has put a moratorium on building new homes; Kansas has stopped watering its crops; the Colorado River is disappearing; and swathes of the American southwest are cracking and sinking from excessive groundwater pumping.
And quantum computing is supposedly next on the horizon. How much water will those servers require?
Trusting the Problem to Invent its Own Solution
It seems deeply cynical—indeed, outright laughable—for slim, slick-haired tech CEOs to claim that AI itself might engineer a solution to this problem as Canada burns and lakes dwindle. AI is trained on human ingenuity. If human ingenuity hasn’t yet managed to squeeze more water out of the bedrock, what’s AI going to suggest? Turning itself off?
Hopefully it comes to such conclusions before all the aquifers run dry.
And all this for a technology that’s unreliable at best. Although generative AI has eliminated thousands of jobs for writers, artists, assistants and programmers, it can’t correctly determine how many ‘m’s appear in the name “Canada” or realize that losing 129 lbs would be deadly for a 130-lb woman. (An eating disorder helpline had to unplug its AI chatbot after it gave dieting advice to clients.) AI also hallucinates, producing entirely false information, and—contrary to layoffs in the creative industry—can only be trained on human-created content. When trained on AI-generated content, the model goes haywire and collapses.
The cynicism that pervades our society, however, is arguably worse. “Yes,” some will admit with a weary sigh, “AI is draining our resources and will probably kill us all. But if you don’t learn how to use it, you’ll be seen as a dinosaur and won’t get a job.”
The cultural posture for the last decade or so has been to adopt whatever new technology rolls down the pipe so we can be the fastest, most efficient, most competitive society on the planet. But when that technology threatens our very existence, adopting it becomes a moral issue.
What will we do when the water runs out? I say “when” because, if our actions on climate change are any indication, we will do nothing to stop it.
