AI Is Not the Only Thing Worth Worrying About

Slop is real. So are the real problems that should not disappear behind it

There is a lot of bad AI work online. There are fake photos passed off as real, low-effort articles stuffed with generic filler, stolen-looking artwork, spammy videos, and people using a chatbot as a substitute for having a thought. Calling that out is fair. "AI slop" is a useful label when the work is deceptive, careless, or made in bulk without any human judgment behind it.

But the fact that slop exists does not mean every use of AI is slop. A tool can be used badly, lazily, creatively, or responsibly. Search engines did not eliminate research. Cameras did not eliminate photography. Spreadsheets did not eliminate accounting. They changed the work, created new ways to cut corners, and also made good work faster and more accessible. AI is no different in that respect.

The conversation gets worse when AI becomes a catch-all villain that absorbs every ounce of public attention. It is easier to rage at a strange AI image on a social feed than it is to follow a city water plan, question an industrial permit, understand hydraulic fracturing, demand responsible data-center development, or hold elected officials and large companies to account for their actual decisions. Those issues are harder, slower, and less satisfying to argue about in a comment section. They are also the issues with the most direct effect on people’s lives.

That does not mean AI should get a pass. It means we should apply the same standard to AI that we should apply everywhere else: ask what is being built, who benefits, what resources it uses, what it replaces, who bears the cost, and whether the claims being made can be verified.

Water deserves scrutiny, not slogan math

Water use is a real concern, especially in dry places such as Arizona. AI infrastructure can require a great deal of electricity and, depending on the cooling system, water. The International Energy Agency expects global data-center electricity demand to keep rising rapidly, which makes efficiency, grid planning, and siting decisions important rather than optional.

At the same time, public arguments often mix together several very different measurements: water withdrawal, water consumption, water reuse, and water that returns to a sewer, river, or aquifer after treatment. Those are not interchangeable.

The U.S. Geological Survey uses the term consumptive use for water that is evaporated, incorporated into a product, consumed by people or animals, or otherwise does not return to local surface water or groundwater for immediate reuse. It is not destroyed. It remains part of Earth’s water cycle. But if water evaporates from a cooling tower in Arizona, it may not fall again as rain in the same watershed, in the same season, or in a form that Tucson can rely on. That local loss is the point of the word “consumed.”

Some of the water used by industrial facilities is not consumed at all. It can be treated, reused on site, discharged to a wastewater system, or recharged into an aquifer. A well-designed facility can reduce its demand for fresh water substantially through reclaimed-water use, dry cooling, and closed-loop systems. But a closed-loop system is not magic: when a cooling tower is used, heat is often removed by evaporating water. The facility must add makeup water to replace evaporation and discharge some mineral-heavy water, called blowdown, to keep the system from scaling up. The Department of Energy explains that evaporation is the main source of cooling-tower water use.

That is why the right question is not simply, “Does this place use water?” Every significant facility does. The better questions are:

How much water is withdrawn, and from where?
How much is actually consumed locally?
How much is reused, treated, or returned to the watershed?
Is it drinking water, groundwater, surface water, or reclaimed wastewater?
Is the project located in an already water-stressed area?
Are its operators reporting site-level numbers openly enough for the public to check them?

The popular 450,000-gallon comparison needs context

Google stated that in 2021 its average data center consumed approximately 450,000 gallons of water per day. That is a serious number, and it should not be waved away. It is also not a universal figure for every data center, every AI service, or every new cooling design. Water use depends on the facility’s size, climate, workload, cooling method, energy source, and whether it uses reclaimed water.

It is also important not to describe that 450,000 gallons as if the same tank of water is endlessly being used and then suddenly disappearing. Cooling systems circulate water internally, and efficient facilities can reuse some water. However, Google used the word consumed for that figure, so it should not be presented as though all 450,000 gallons are automatically available for local reuse at the end of the day. Without site-specific data, we cannot honestly calculate how much of that amount was newly withdrawn, recycled on site, returned after treatment, or evaporated. That gap is an argument for better disclosure, not for making assumptions in either direction.

Now compare that number with car manufacturing. Toyota’s North American environmental metrics list 887 gallons of water withdrawn per vehicle produced for fiscal year 2025. At a hypothetical output of 1,000 vehicles per day, that works out to about 887,000 gallons withdrawn per day. It shows why public concern should not narrow itself to data centers while treating older industries as invisible.

But this is where honesty matters. The Toyota number is a withdrawal figure per vehicle, while Google described its number as consumption per data center. Those are different measures with different reporting boundaries. So it is reasonable to say that auto manufacturing can operate at a comparable or larger water scale, and that it deserves the same scrutiny. It is not accurate to use these figures alone to prove that “car manufacturing uses exactly twice as much water.” The arithmetic is simple. The comparison is not.

The bigger point survives the correction: data centers are not the only major industrial water users in Arizona or anywhere else. Automotive manufacturing, semiconductor fabrication, power generation, mining, agriculture, construction, and conventional manufacturing all draw on water supplies. In Arizona, chip fabs are particularly relevant because making semiconductors requires extremely pure water and can involve very large volumes. AI’s real footprint includes not only the visible data center, but also the electricity and hardware supply chains behind it.

Criticism should be aimed at the failure, not the tool itself

There are legitimate reasons to demand rules for AI. Nonconsensual deepfakes, fraud, impersonation, automated spam, uncredited copying, and deceptive political content are problems. So is a company hiding resource use behind vague sustainability language. None of those problems are solved by pretending that every person who uses AI to brainstorm, organize notes, learn a subject, clean up a rough draft, code a small tool, or make art is doing something dishonest.

The useful dividing line is not “AI or no AI.” It is whether a person is using the tool with responsibility and judgment.

If the output lies about being real, steals someone’s work, spreads misinformation, replaces thought with filler, or wastes public resources without transparency, criticize it. If the tool helps someone communicate more clearly, learn faster, access creative work, reduce repetitive labor, or turn a rough idea into something they can actually build, then it can be a benefit.

The same standard should apply to companies building the infrastructure. They should be expected to publish meaningful water and energy data, use reclaimed water or low-water cooling where practical, avoid dumping disproportionate costs on drought-stricken communities, and prove that their promised jobs and economic benefits are real. That is not anti-AI. It is basic accountability.

We can worry about more than one thing

The honest position is not that AI has no costs. It does. The honest position is also not that AI is uniquely monstrous while every other water-intensive industry, every public-policy failure, every unchecked permit, and every political distraction gets ignored.

People should be able to reject slop without rejecting every useful tool. They should be able to demand water accountability from data centers without acting as if a gallon used by a chip fab, factory, power plant, or agricultural operation suddenly does not matter. And they should be able to keep their attention on the broader questions of public oversight, corporate power, infrastructure, affordability, and leadership instead of letting one flashy technology dominate the entire conversation.

AI is not a replacement for human judgment. It is a reason to use more of it.

Sources

U.S. Geological Survey, Water Use in the United States and Water-Use Terminology
U.S. Department of Energy, Cooling Water Efficiency Opportunities for Federal Data Centers
Google, Our commitment to climate-conscious data center cooling
Toyota Motor North America, Environmental metrics table
International Energy Agency, Energy demand from AI