This is actually such a misunderstood topic and after spending already over 100 hours looking into water as a theme I have two points I want to bring to light and argue against as I see them brought up every time Water is introduced as an investment, but before we do that lets start with two definitions
Withdrawal is water removed from a source
Consumption is the fraction of that water that does not return to the local watershed, overwhelmingly because it evaporated
Evaporative cooling doesn’t lose water as a side effect, the loss is the cooling, the heat leaves with the vapor, the cooling effect is the latent heat of vaporization being carried off
Now back to the points I want to argue, “Water doesn’t disappear”
Think about how your own body sheds heat, when you sweat through a workout, that water doesn’t vanish it evaporates into the air around you, every molecule still in the room and yet you’re dehydrated, and “the water’s still in the building” does nothing for you
You have to go drink more, nobody would tell a marathoner he can’t be dehydrated because his sweat is floating around the stadium, a cooling tower is doing the identical thing at industrial scale
Second bad take I see is “So they’ll just go closed loop”, actually not bad take but very lazy, allow me to explain
Cooling happens in two stages, server level cooling moves heat off the chip, air, direct to chip cold plates, or immersion and this stage consumes essentially no water
Then facility level cooling rejects that heat to the outside environment, the densification of AI racks is what’s forcing the shift to liquid at the server stage direct to chip removes 70–80% of the heat load right at the chip ,but that’s a transport problem, not a water problem
The facility loop ultimately rejects its heat through a condenser, a dry cooler, or a cooling tower, “We went liquid cooled” answers the first question and says nothing about the second, a rack of immersion cooled GPUs can still have its heat dumped by an evaporative tower burning through millions of gallons…,conflating “liquid cooling” with “waterless” is not a serious comparison
Okay so now this is going to get technical so stay with me
The real question we should be asking is never air vs. liquid, It’s how the facility rejects heat
Energy is not the cost the operator is trying to minimize at the margin, power is the binding constraint of the entire buildout, interconnection queues run years; the scarce input is megawatts, not gallons and not dollars
During peak summer conditions, evaporative or evaporatively assisted coolers use 10–35% less electricity than equivalent air cooled systems because evaporation sheds heat that a dry cooler can only fight with brute fan power
In a world where every megawatt is spoken for, that saving is most valuable exactly when it’s scarcest like the hot afternoon when the grid is also at peak, so the binding power constraint biases operators toward spending water, not away from it
A dry cooler can only reject heat down toward the dry bulb temperature, while an evaporative system rejects toward the wet bulb temperature
Evaporative cooling still works even when the dry bulb exceeds the loop setpoint, as long as the wet bulb sits a few degrees below it, when it’s 43°C in Phoenix and your liquid loop wants 40°C, a dry cooler is nearly useless
The air is hotter than your setpoint while an evaporative tower runs fine because the wet bulb might be 22°C
The drier the air, the wider the gap between dry bulb and wet bulb, so evaporative cooling is most effective exactly where the air is driest, which is exactly where water is scarcest
There’s no version of this buildout that doesn’t run through water, every cooling design just decides how much, and where
The money in water is in the merchant tier, the marginal gallon that clears at scarcity pricing when the system is tight, not challenging the future of cooling or dethroning bottle necks
I was thinking, wouldn't water related stocks go brrr in the next years due to the insane data centers water usage global warming population growing? I feel like these 3 components put together are the perfect recipe for a water shortage in the future, am I missing something?