Okay, while I do suspect the 16oz number is pulled from someone's ass (or just very old, many efficiency improvements ago), there are a couple of things to consider:

a) A single B300 won't even run frontier models. It only has 288 GB of HBM. The DGX B300 server has 2.1 TB, which I suspect is not enough either (running Qwen 3.8 Max lossless (BF16) requires 4.9 TB just to load the model). But these are run in clusters anyway and any single user's prompt only uses a tiny percentage of the compute from tons of servers. OpenAI and Anthropic are going for size with their models so the most expensive ones probably need even more than Qwen.

b) While the GPUs themselves are directly cooled by closed loops, the water inside those closed loops is often cooled by evaporative coolers, or even AC and evaporative cooling in tandem (that allows the evaporation to happen outside the datacenter itself). That's because evaporative cooling is the most energy efficient form of cooling.

There are companies offering very complex heat systems that take advantage of multiple forms of cooling because they're making 100+ kW racks these days and talking of 500-1000 kW racks in the future. Ridiculous numbers, honestly.

This is from some company's website that offers cooling solutions: Notice the outside "cooling tower" - those can be dry or wet. In a hot and dry climate, you'd want wet. It'll cool really well. In a colder and more humid climate, dry might work better. They also optionally let you heat the aquifer apparently, which is bound to cause issues for everyone living nearby over time.

You're talking about closed water cooling systems, which are ethical and more expensive.

Now can you guess how many data centers use those?

I believe the paper was looking at total footprint including building and running data centers with open evaporative cooling, and the water "used" by civilian power stations.

I don't really see more water in the air as the most pressing environmental issue around data centers. That's better than the water being destroyed or contaminated.

The main environmental hazard of data centers is the rapid expansion of them. The running of them for months on end in the training of models... And above all else, the fact billionaires have chosen to power them in the least environmentally friendly ways (gas and disel):

https://duckduckgo.com/?q=dirty+power+ai+data+centers+powered+by

There's also data centers using closed loop systems, they don't lose water

I think you're right about the number being nonsense, but water not being destroyed doesn't mean you haven't converted potable water to non-potable water, and it is still important at the actual scale of water that data centers are using for cooling to the localities they operate in.

The extremely dumb thing here is that the water use is totally fine. California alone uses ~11 TRILLION gallons of water per year for farming. And so what, that water cycle returns via precipitation (assuming we don't go so hard to fuck up the aquifers, but you know what I mean). The water is fine. Plus that farm water runoff is even kinda hazardous, with pesticides etc. It's still fine

This is a totally recyclable resource. You can probably use it to water plants after it's been through the data center, or it may have just evaporated away as part of the cooling process

Meanwhile, the obvious dart to throw is the fact that the power need is absolutely ridiculous and only growing, leading to every major tech company investing extensively in energy tech. Eventually they will become a new generation of utilities that supply power (and compute) to everyone. So the opportunity is to force them to reinvent energy in a clean way rather than, say, burning an utter shitton of oil to do it