This is not an isolated case in the United States, a country that concentrates
approximately 30% of all the data centers in the world. Arizona, Utah and South
Carolina are well aware of the insatiable thirst of this type of infrastructure.
They are also familiar with it in the Netherlands, where Microsoft was involved
in a scandal last year when news broke out that one of its facilities consumed
four times more water than declared in a context of drought. Or in Germany,
where Brandenburg authorities denied Google permission to build a data center in
the region, as a Tesla gigafactory was already consuming too much water.
Data centers reuse water by recirculating the same water through their cooling systems multiple times. According to Google, this practice saves up to 50% of water when compared with “once-through” cooling systems.
However, eventually this reused water needs to be replaced with new water, due to the risk of scale formation or once the conductivity of the water is too high.
The need for new water results from the build-up of scale-forming minerals in the reused water – such as calcium, magnesium, and silica, which become concentrated over multiple evaporative cooling cycles.
Ultimately, the spent data center cooling water can be treated and reused (or recycled) for other purposes such as irrigation or toilet flushing.
But it doesn’t have to be drinking water. Nuclear power plants, for example, often use 3 cooling circuits. The first two are closed loops, in order to avoid the release of radioactive nuclides. The coolant is condensed (using heat exchangers with the next circuit) and recirculated. The last circuit is often just river water or similar that’s thrown out after use. Even the evaporated water isn’t an issue, since it will fall back as rain somewhere. The atmosphere has a limited capacity to hold water vapor.
My real concern with AI isn’t water at all. It’s the energy usage. Water (not drinking water) is renewable. The bulk of the electric power supply is not. Perhaps someday, there will be technology to do the training with much less power. But today it’s unsustainable. But the big players will keep doing it, since they make money off of it. The incentives are just as perverse as with the crypto mining industry. And just like crypto, AI is headed in a way where a few rich players have all the edge to become even richer, at the expense of regular folks.
If it make you feel better, they could do it all with one closed loop and just use the ground as the sink.
I don’t how that should make me feel any better 😀 . But I don’t know if ground is a good enough sink for that.
They could then power it entirely off solar panels and Iron air or sodium batteries.
I don’t think they’re going to consider renewables for cooling alone when the entire operation needs enormous amounts of power that cannot be satisfied by renewables.
They’d cool their servers with the blood of children if it saved them money.
No, but that is generally what buildings are piped for. Using a simple evaporative cooler and municipal water is the easiest and likely cheapest option.
Is there a reason why that water can’t be reused?
I always assumed it was for cooling
Data centers reuse water by recirculating the same water through their cooling systems multiple times. According to Google, this practice saves up to 50% of water when compared with “once-through” cooling systems.
However, eventually this reused water needs to be replaced with new water, due to the risk of scale formation or once the conductivity of the water is too high.
The need for new water results from the build-up of scale-forming minerals in the reused water – such as calcium, magnesium, and silica, which become concentrated over multiple evaporative cooling cycles.
Ultimately, the spent data center cooling water can be treated and reused (or recycled) for other purposes such as irrigation or toilet flushing.
https://dgtlinfra.com/data-center-water-usage/#:~:text=What Happens to Water Used,a local wastewater treatment plant.
Evaporative coolers release the vapor into the atmosphere.
But it doesn’t have to be drinking water. Nuclear power plants, for example, often use 3 cooling circuits. The first two are closed loops, in order to avoid the release of radioactive nuclides. The coolant is condensed (using heat exchangers with the next circuit) and recirculated. The last circuit is often just river water or similar that’s thrown out after use. Even the evaporated water isn’t an issue, since it will fall back as rain somewhere. The atmosphere has a limited capacity to hold water vapor.
My real concern with AI isn’t water at all. It’s the energy usage. Water (not drinking water) is renewable. The bulk of the electric power supply is not. Perhaps someday, there will be technology to do the training with much less power. But today it’s unsustainable. But the big players will keep doing it, since they make money off of it. The incentives are just as perverse as with the crypto mining industry. And just like crypto, AI is headed in a way where a few rich players have all the edge to become even richer, at the expense of regular folks.
deleted by creator
I don’t how that should make me feel any better 😀 . But I don’t know if ground is a good enough sink for that.
I don’t think they’re going to consider renewables for cooling alone when the entire operation needs enormous amounts of power that cannot be satisfied by renewables.
Amen to that! Like I said - perverse incentives.
deleted by creator
No, but that is generally what buildings are piped for. Using a simple evaporative cooler and municipal water is the easiest and likely cheapest option.