"Two-phase immersion cooling is not dead. ZutaCore and Chemours are both pushing toward PFAS-free solutions that could revive the segment. But the timeline for commercial-scale availability of those alternatives stretches through 2026 and into 2027. The EU regulatory decision could land before the alternative fluids do. And the hyperscalers who were the most likely buyers at scale have already moved on."
Feels analogous to looking for cold fusion because of the downsides of fission.
I didn't get a significant sense of loss from this article though. Especially given the downsides of PFAS.
It took me about a second to realize the link took you to a list of articles. It took another second to realize the article referenced was second in the list.
> 3M did not make this call because they found a better product. They made it because they were staring down over 4,000 lawsuits and a $12.5 billion settlement with more than 11,000 U.S. public water systems alleging PFAS contamination in drinking water. The settlement received final court approval in March 2024. Payments stretch over 13 years.
> Who Got Hurt, Who Didn't
It seems to have had an massive impact, and article also includes a "who got hurt" but there is zero numbers about the number of people actually impacted by this catastrophe? I'm guessing the blogs focus might be on businesses, but considering this might be spawning something of a health crisis in the affected areas, maybe at least a mention of the humans involved here would make sense.
At least where I live the PFAS disaster is so widely known and publicized that I would say it's fair to leave it out of scope for this article. Oh and "who got hurt" - basically everyone worldwide for the foreseeable future.
> At least where I live the PFAS disaster is so widely known and publicized that I would say it's fair to leave it out of scope for this article
Doesn't that kind of assume that I and everyone too also been impacted, so we should have read about it in our local news? I don't think 3M's PFAS disaster ever been mentioned in either my country's newspaper, nor my local paper, my first time reading about it here, so would be nice if the article didn't make such assumptions.
I would think, but maybe not. In the US, the story was enormously reported on, perhaps second only to the Epstein files, with a long tail that still persists now. They are also called "forever chemicals", if you've heard that term.
Many municipalities across the country were/are forced to upgrade their water filtration systems - a huge cost, possibly too little and too late. I know many other countries are taking action too, but I don't know how it compares to how the US responded.
> They are also called "forever chemicals", if you've heard that term.
Yes, of course I've heard about PFAS before, but not that data centers were polluting public water systems with PFAS, did you mean earlier that you've read stories about PFAS in general in your papers? That'd put your previous comment in another light, I thought you specifically talking about the "more than 11,000 U.S. public water systems alleging PFAS contamination" part, not PFAS' in general.
I've never understood one vital thing - if PFAS is by nature totally inert and unreactive, how is it harmful? If you drank a glass of the stuff, what would happen?
As I understand it they discovered a long chain molecule which was highly inert and wouldn’t stick to anything. Which was a useful feature but you know makes it hard to attach to anything. So they created a similar smaller chain molecule which had a reactive tip but was still super stable. Unfortunately it’s also a bit amino acid like. So we ended up with a molecule which is very durable and accumulates in living things.
Then of course we produced it at industrial scale for decades flooding the entire planet with this stuff.
For example PTFE is a large molecule with strong bonds, and as a consequence isn't very reactive and likely safe.
On the other hand, perfluoroalkyls such as PFOA have the same shape as fatty acids, so they bind to the same places such as in the liver, which makes them grave health hazards.
Many precursors used for making PFAS are also toxic, so for example, even if PTFE is safe, manufacturing it isn't.
The bioaccumulate part I understand, for the body to eliminate something it has to bind to it somehow. Tough to do if the chemical won't react with anything.
I'm not sure what "biologically inert" means specifically. Are you saying there are biological chemicals that actually do interact with this stuff? A single example would help me understand.
I don't know the details here for PFAS (and they likely would vary enormously for the different molecules that fall into this broad category). But in general a molecule doesn't have to react to be accumulated. Inert usually means it doesn't react with other substances in a normal environment. It doesn't mean you can't make it react if you add enough energy. For example nitrogen gas is considered inert. Bacteria (or chemical plants) can make it react and produce different nitrogen-containing molecules from it.
Inert doesn't really say anything about toxicity, it's not directly related to that. The opposite is though, pretty much any strongly reactive chemical is dangerous or toxic in some way since it will react with stuff humans are made from.
With PFAS the inert example is also usually Teflon. That is also a solid polymer, so not many individual molecules. There isn't much you body could do to process a macroscopic chunk of Teflon, so you'd almost certainly just excrete it.
As far as I understand the nomenclature, PFAS covers both the inert final products like Teflon and reactive intermediates, degradation products and reactants. It's a very broad category of chemicals.
My understanding is that the bigger danger is e.g. a Teflon-producing plant than the final Teflon products (assuming the Teflon isn't damaged and heated too much). Because the plant has to handle the reactive ingredients, and those can leak into the environment.
Edit: Nevermind, Wikipedia makes it pretty clear that even the non-broken-down PFAS are totally unsafe, evil things which we knew were dangerous since the 70s and did nothing about until recently
Fluoroalkyl chemicals are only "inert and unreactive" in a relatively narrow sense of "wouldn't catch fire", "don't react with strong acids and bases", and similar.
They are plenty reactive in a sense of interacting with enzymes and other cellular machinery.
Not really accurate. These chemicals are quite unreactive. Precursors from manufacturing waste can be very reactive, but most of the problematic contamination regards the forever chemicals themselves, not precursors. This paper is probably the best scientific review of what is going on in the human body. https://www.sciencedirect.com/science/article/abs/pii/S03043...
Maybe sci-hub has a copy of the full paper. Not sure.
As briefly as possible, and therefore glossing over many many details, the toxic effects are mainly due to cell membrane perturbation, cell membrane transport disruption, and binding to hydrophobic protein cavities (thus disrupting the usual function of these cavities).
PFAS used to be considered totally inert but later research showed correlations between bad health effects and higher concentrations of PFOA and PFOS.
3M and DuPont knew since the 1970s and suppressed the information, not dissimilar to how tobacco and oil industries created disinformation about externalities.
Making PFAS and having to dispose of byproducts is the nasty part as far as I understand. There is also some kind of reaction that can happen where it will off gas nasty enough stuff to kill your pet bird if you overheat your pan.
TFA points out that single-phase immersion systems work fine without PFAS. This change affects two-phase systems that rely on the coolant evaporating at a low temperature (60C) and recondensing.
Florinert can be formulated for either one-phase or two depending on boiling point. Mineral oil is only suitable for single-phase because you cannot deep fry your CPU (it boils at 200C)
In cold climate markets with a lot of fiber or other very fast access like very good 5G, why don’t AI companies or their cloud vendors market home or building scale heating compute nodes?
They could be built so that they exhaust waste heat into the HVAC system in winter and then switch to an outside piped radiator in the summer or something similar.
End users wouldn’t buy it. They’d make some kind of deal where one is installed and they pay less for heat and the extra electricity is paid by the compute operator via a separate meter. So the DC operator gets cooling that is (averaged over the year) almost free since they are basically reselling the heat for half the year or more.
For individual homes it might be unwieldy to manage a bunch of small units, so apartment/condo blocks and businesses might make more sense for installation. They could be colocated with building HVAC.
I guess economics depends on what percentage of DC cost is power or water for cooling.
It's a relatively new thing but over here, in a cold climate, data centers are now often built so they are connected to district heating networks and the waste heat is utilized.
Every use case isn’t security critical. There’s a lot of AI used to render silly images and films, generate marketing copy, play games, proofread social media posts, etc.
You solve this by charging a little more for security certified AI that runs in a secure DC.
You can also easily randomize AI work load distribution, so individual nodes don’t get all queries for a whole coherent conversation or project. It makes an attack less valuable.
Apple has done some research on blinding and obfuscating AI too.
Feels analogous to looking for cold fusion because of the downsides of fission.
I didn't get a significant sense of loss from this article though. Especially given the downsides of PFAS.
> Who Got Hurt, Who Didn't
It seems to have had an massive impact, and article also includes a "who got hurt" but there is zero numbers about the number of people actually impacted by this catastrophe? I'm guessing the blogs focus might be on businesses, but considering this might be spawning something of a health crisis in the affected areas, maybe at least a mention of the humans involved here would make sense.
Doesn't that kind of assume that I and everyone too also been impacted, so we should have read about it in our local news? I don't think 3M's PFAS disaster ever been mentioned in either my country's newspaper, nor my local paper, my first time reading about it here, so would be nice if the article didn't make such assumptions.
Many municipalities across the country were/are forced to upgrade their water filtration systems - a huge cost, possibly too little and too late. I know many other countries are taking action too, but I don't know how it compares to how the US responded.
Yes, of course I've heard about PFAS before, but not that data centers were polluting public water systems with PFAS, did you mean earlier that you've read stories about PFAS in general in your papers? That'd put your previous comment in another light, I thought you specifically talking about the "more than 11,000 U.S. public water systems alleging PFAS contamination" part, not PFAS' in general.
Then of course we produced it at industrial scale for decades flooding the entire planet with this stuff.
For example PTFE is a large molecule with strong bonds, and as a consequence isn't very reactive and likely safe.
On the other hand, perfluoroalkyls such as PFOA have the same shape as fatty acids, so they bind to the same places such as in the liver, which makes them grave health hazards.
Many precursors used for making PFAS are also toxic, so for example, even if PTFE is safe, manufacturing it isn't.
I'm not sure what "biologically inert" means specifically. Are you saying there are biological chemicals that actually do interact with this stuff? A single example would help me understand.
Inert doesn't really say anything about toxicity, it's not directly related to that. The opposite is though, pretty much any strongly reactive chemical is dangerous or toxic in some way since it will react with stuff humans are made from.
With PFAS the inert example is also usually Teflon. That is also a solid polymer, so not many individual molecules. There isn't much you body could do to process a macroscopic chunk of Teflon, so you'd almost certainly just excrete it.
My understanding is that the bigger danger is e.g. a Teflon-producing plant than the final Teflon products (assuming the Teflon isn't damaged and heated too much). Because the plant has to handle the reactive ingredients, and those can leak into the environment.
Edit: Nevermind, Wikipedia makes it pretty clear that even the non-broken-down PFAS are totally unsafe, evil things which we knew were dangerous since the 70s and did nothing about until recently
They are plenty reactive in a sense of interacting with enzymes and other cellular machinery.
Maybe sci-hub has a copy of the full paper. Not sure.
As briefly as possible, and therefore glossing over many many details, the toxic effects are mainly due to cell membrane perturbation, cell membrane transport disruption, and binding to hydrophobic protein cavities (thus disrupting the usual function of these cavities).
3M and DuPont knew since the 1970s and suppressed the information, not dissimilar to how tobacco and oil industries created disinformation about externalities.
I think it was advertised as very water proof, like water would pearl on it.
It's probably full of PFAS, no idea if it has been leeching PFAS, but I know it's not very waterproof anymore, so that might be a worrying clue.
Florinert can be formulated for either one-phase or two depending on boiling point. Mineral oil is only suitable for single-phase because you cannot deep fry your CPU (it boils at 200C)
They could be built so that they exhaust waste heat into the HVAC system in winter and then switch to an outside piped radiator in the summer or something similar.
End users wouldn’t buy it. They’d make some kind of deal where one is installed and they pay less for heat and the extra electricity is paid by the compute operator via a separate meter. So the DC operator gets cooling that is (averaged over the year) almost free since they are basically reselling the heat for half the year or more.
For individual homes it might be unwieldy to manage a bunch of small units, so apartment/condo blocks and businesses might make more sense for installation. They could be colocated with building HVAC.
I guess economics depends on what percentage of DC cost is power or water for cooling.
You solve this by charging a little more for security certified AI that runs in a secure DC.
You can also easily randomize AI work load distribution, so individual nodes don’t get all queries for a whole coherent conversation or project. It makes an attack less valuable.
Apple has done some research on blinding and obfuscating AI too.