AI Data Center Contamination by Larry Good

This is going to be a two-part talk. First I’m going to address “AI Datacenter Water Pollution” and because that will already be long enough, I’ll do a second talk on water consumption. I may have to do a third on electricity consumption and the planning processes for “Large Load” consumers and their effect on grid infrastructure. That’s another area where I am not the definitive expert, but I’ve picked up more insight than the average person, merely by being in the industry and in a risk management capacity, looking at future impacts and technical planning needs.

For background: I work in Information Technology, and most of my career has now been in large infrastructure. Internet banking, large public-school districts, telecommunications, DoD support (including 5 years forward deployed to combat zones) and for the last 13 years, supporting the bulk electric system (‘the grid’) at a large utility.

I don’t work in AI datacenters specifically, but I’ve been doing data center operations, planning, engineering, and cybersecurity support for quite a few years. Running this big chunk of the national bulk electric system grid literally cannot be done effectively and efficiently without computers. From the physical plant perspective, it doesn’t matter what purpose the computing resources are being used for. I’ve done a lot of capacity planning for physical space, HVAC, and continuous & backup power requirements for IT installations. I’m not a civil engineer, but I’ve had to interface and collaborate with many over the years, and for budgeting and resourcing purposes I have picked up a general knowledge of the necessary infrastructure considerations as it pertains to large, multi-year, multi-million-dollar projects.

Currently, the public zeitgeist is all atwitter with panic over the water supply being sucked up into the ‘AI Datacenter infrastructure that will enslave us for 1000 years, and poison the earth forever’. Look, I’m not a huge fan of AI but let’s not be luddites, either.

And as part of the “drank from the garden hose” generation, I have a lot of disdain for the hype about water pollution in this context. It’s (mostly) bullshit.

Singling out this ONE type of commercial/industrial application simply isn’t rational. A large part of it is (as so many things on the Internet currently are, especially with AI making it easier and faster) nation-state and other organized propaganda farms generating FUD for their own purposes. Like those other nation states “winning the AI race”. That’s my cyber-security/critical infrastructure protection specialist hat I’m talking through there.

Yes, data centers in general use more cooling *on the data center floor* than the average office building (per floor, at least. Many office complexes are WAY bigger than most datacenter facilities), but less than the average industrial plant. We have a more concentrated electric usage on floors with server racks (than cubicles), but we don’t have nearly the personnel usage of environmental controls, water fountains, flushing toilets, and inefficient heating/cooling/lighting needs for human comfort and large open spaces. This difference in water usage is significant. Office cubicle farms really require a lot of juice too, with all the individual machines, monitors, peripherals, lights etc. Human density on the data center floor is tiny compared to cubicle farms in office buildings. Traffic is low, with all kinds of knock-on effects like we don’t need janitors vacuuming carpeted floors daily. We don’t need lights on all the time.

Yes, the water flowing through the cooling system is technically industrial wastewater. As one supremely over-confident person told me: “once it leaves the distribution system, it isn’t drinking water anymore.” True: water that has been used in a cooling system is *deemed* non-potable by standard building codes and conservative safety practice, not by tested mineral content.

It is non-potable water. Which does not mean you would be poisoned by drinking it directly from an effluent pipe. It’s a definitional thing. If your kid is homeschooled at the age when your neighbors’ kid is in kindergarten, the neighbor kid is a kindergartner and yours is not. By definition.

Water in your toilet tank (not talking about the bowl- not yet) is the same water coming out of your kitchen tap. Is that water in the toilet tank drinkable? Would you go scoop a glass of water out to drink? Would you drink it confidently in an emergency? (I would). Water coming out of your kitchen tap into your kitchen sink; would you scoop a glass of water out of your filled sink and drink it? What about using your bathtub for storing water in advance of a predicted storm or natural disaster? Would you drink that? I think for most people it depends on need, and they can acknowledge that the reaction of disgust is not rational, it’s emotional and could be overcome by a risk-based evaluation of cost/benefit.

What about water in a freshly-cleaned and sanitized toilet BOWL, that hasn’t been excreted into. Is that water “*polluted*? I wouldn’t want to drink it. But I would have no problem putting it on my garden.

All of these scenarios provide a risk/reward calculation that a person has to figure for themselves, and all deal with less than perfect water quality, but none of them are absolute, and none of this water is “polluted”. But here’s another thing- your municipal tap water isn’t perfect either. You just accept that it is “good enough” and never think about it unless it hits the headlines with a “boil water” advisory. We don’t even know what the perfect mix and amount of minerals in water truly is, or even that there is a perfect mix for all populations under all conditions. Although I’m sure the makers of Brawndo would tell you their sportsball drink is the perfect mix: because it has electrolytes (sodium chloride, iron, magnesium, calcium, etc.).

But yes, industrial/commercial buildings generate water that is not recommended for drinking directly.

Water in cooling towers evaporates. It doesn’t get treated by industrial process. It just rains down. The remnant may have some higher levels of metal & minerals in it than the treated inputs from the municipal supply. So does the water going INTO the municipal supply before it is treated. The municipal supply typically ADDS some other chemicals- chlorine, fluorine. By code, the remnant water from g evaporative towers can be discharged into normal sanitary drains, even if the mineral content is slightly more concentrated. That means decades of engineering and testing has deemed it safe. It is by any rational standard, MUCH less dangerous than your toilet water which gets discharged into the same sort of drain.

Most of the time, the municipal water supply is used by default because it’s a convenient source of relatively clean water to start with. It may even be REQUIRED by local ordinance and city/county planners. Sometimes it still has to be filtered again at the industrial site prior to use. Some industrial processes use de-ionized water. It can’t even have the normal mineral content that humans NEED in our water (trace minerals) and cannot tolerate the additives like chlorine and fluoride used for sterilization and tooth treatment.

Depending on the municipal supply’s mineral content (varies wildly across the USA, as long as it is within EPA tolerances), cost, and delivery availability, a closed loop system might find they don’t want the municipality involved at all. They might have water trucked in, or tap a natural supply from roof run-off or natural water source, and filter it themselves. In that case, the water they discharge could potentially be “cleaner” than what they take in.

In no case are they going to get permitting to release contaminated water that is a risk to the public. And EVERY building must go through a building permit process, and EVERY commercial building gets even more scrutiny for water usage, runoff, even plans for stormwater management.  Alexandria Ocasio Cortez was in the news lately: her jars of brown water stunt were supposedly taken from a site where a datacenter was *under construction*, not *in operation*.

“Rep. Alexandria Ocasio-Cortez displayed two bottles of dirty water that she claimed came from Morgan County, GA, where a Meta data center is being built.”

(Source: https://www.yahoo.com/…/aoc-pulls-two-jars-brown…)

“Being Built.”

AOC didn’t have *data center effluent*. If what she had even came from a municipal water supply (doubtful), she had evidence of local government corruption and incompetence (common), not capitalist exploitation.

Some cooling systems might (not definitely, but may) use piping that may not be certified for drinking water. Perhaps it has soldered connections that aren’t rated for consumption. But closed loop systems aren’t just discharging it. And glycol is expensive. You don’t discharge “cooling chemicals” and refrigerant, you KEEP IT in closed loop systems. Data centers are hardly the only, or even primary user of this tech. EVERY OFFICE BUILDING constructed in the last century, every movie theater, every grocery store, feed store, department store and government building uses it too.

‘Contaminants’ are not necessarily persistent poisonous toxins, but may be anything not necessary for human health in drinking water. Could be irrelevant to human health. Or too much of something you do need: too much salt, too much calcium or iron, and BTW there are actually allowable levels even for arsenic coming into your drinking water from primary sources like aquifers. Parts of the USA have always had arsenic levels.  Right now, there is a democrat effort underway to block testing for birth control hormones and abortifacients in municipal water supplies, because they say “it’s a witch hunt to demonize and ban birth control, and couldn’t possibly be at a level that is harmful.” (Weaponizing Water: How the Campaign Against Medication Abortion Co-opts Environmental Policy | Guttmacher Institute)

Industrial standards already exist for what level of contaminant are allowed to be discharged back into the municipal sewers, and the municipal sewage system then treats before discharging into natural bodies of water or returning to the drinking water distribution system. When it exceeds these levels, it must be treated on site first. These standards even extend to *water temperature* before being discharged into natural bodies of water- but such discharge is almost never allowed in the permitting process anyway.

In the cases where water is contaminated at an industrial building, it simply gets piped across to the building where they process it, and piped back into the cooling loop. Doubly so for silicon fabs. Most industrial processes and any chemical plants do this all day, every day, with far more dangerous “pollutants” than will ever leach out of a modern air conditioner’s guts.

This is a solved problem.

Larry Good, Author at Ordinary Times
Facebook: Lawrence Good, Scribbler