Byte BungalowByte Bungalow

← Stories

Someone Asked If You Can Actually Buy the Underwater-Data-Center Boom. I Went Digging.

A commenter left the one question a normal person actually cares about: are there any publicly traded companies in this? Fair. You watch a clip about servers sitting on the ocean floor or coolant running through veins etched into a chip, and part of your brain goes, wait, can I own a piece of that? So I spent an afternoon chasing it down. And the honest surprise is that the stuff that looks the coolest is mostly the stuff you can't buy — while the money you can actually touch is hiding one layer down, in the plumbing nobody makes a video about.

So can you buy the underwater data center?

That was my first stop, because the seafloor pods are the thing everybody's seen. And here's where it got weird: the most famous one is already dead. Microsoft ran Project Natick, sank a sealed tube of servers off Scotland, and it worked — fewer failures than the ones on land. Then in 2024 the head of their cloud operations group flatly said she isn't building subsea data centers anywhere in the world. They kept it as a research toy and walked away.

So who's still doing it underwater? China. There's a commercial pod running off Hainan Island and a bigger seawater-cooled project going in near Shanghai. But that's a state-backed effort, not a ticker sitting on the Nasdaq waiting for your order. So the underwater angle, the thing that pulled you in? There's basically no clean way to buy it. You'd be buying a research project somebody already shelved, or a Chinese government program you can't reach.

The veins in the chip belong to a startup you can't touch

The other headline is the one where they stop cooling the outside of the chip and start running coolant through tiny channels carved into the silicon itself — like giving the processor its own bloodstream. Microsoft showed this off in the fall of 2025, and the number that made me sit up: it pulled heat away up to three times better than the cold plates everyone uses now, and knocked the peak temperature spike down by roughly two-thirds. That's the difference between a chip that throttles and one that just keeps running flat out.

So naturally I looked up who makes it. Turns out Microsoft built it with a small Swiss startup. Private. No shares. And this is the pattern that kept repeating: the genuinely new cooling tech tends to live inside a tiny private company partnered with a giant. You can go buy the giant, sure — but then you own all of Microsoft, and the cooling is a rounding error in a company mostly made of cloud contracts and Office subscriptions. It's like buying the whole grocery store because you liked one brand of coffee on aisle seven.

Where the public money actually sits: the boring plumbing

This is where the trip turned. Once I stopped hunting for the pod and the magic chip, the publicly traded names showed up fast — they're just one layer down, in the pipes and pumps and racks. The cold plates that sit on the chip. The coolant distribution units that push the fluid around. The manifolds. The people who physically install all of it in a building.

The closest thing to a pure bet is Vertiv, which trades on the NYSE and has bent itself almost entirely toward AI cooling and power. When I looked, their booked-but-not-yet-built order pile was north of fifteen billion dollars — call it a year to a year and a half of work already sitting in the shop before they sell another thing. There's also nVent, which does the liquid-cooling distribution guts, and Comfort Systems USA, which is the crew that actually bolts the cooling and electrical into the data center. None of them make a dramatic video. All of them are things you can buy before lunch. So the answer to the commenter is basically: you can't buy the pond, but you can buy the company that sells the pipes that go into the pond.

The vanishing specialists — and one last warning before you go looking

Here's the twist I didn't expect. The private cooling shops that pop up in these videos — the immersion-tank makers, the cold-plate specialists — keep getting swallowed whole. In the last stretch, Trane agreed to buy LiquidStack, Eaton scooped up Boyd's thermal business, and Ecolab bought CoolIT for billions. So the small pure-play you were trying to find? Half the time you literally can't, because a bigger public company already ate it. Which flips the whole question: don't hunt the startup, watch who's doing the eating.

And one caution, because I ran into it and you will too. Search 'AI liquid cooling stock' and you'll trip over a handful of tiny foreign-listed tickers announcing cooling joint ventures with a lot of exclamation points. Small, thinly traded, and nothing like Vertiv in size. I'm not telling you what to do with your money — just that the flashy press release and the sturdy business aren't always the same company. If you want to actually understand which of these players is doing the real work before you go near any of it, that's the whole reason Byte Bungalow exists. Watch the full breakdown and subscribe on YouTube — we go down these rabbit holes so you're not squinting at a penny stock at midnight wondering if it's real.

Common questions

Are there any publicly traded companies in liquid and underwater data center cooling?
Yes — but not the ones you'd guess from the videos. The underwater pods and the coolant-in-the-chip tech are mostly shelved, foreign, or private. The public names are the plumbing suppliers and installers: Vertiv, nVent, and Comfort Systems USA are the most direct, with bigger diversified players like Eaton, Trane, Carrier, Schneider Electric, and Ecolab now piling in through acquisitions.
Can I invest in the underwater data centers themselves?
Not really. Microsoft ran the famous Project Natick experiment and then shut it down in 2024. The active underwater programs today are largely Chinese state-backed efforts off Hainan and near Shanghai, which aren't something a retail investor can buy shares in. There's no clean public 'underwater data center' stock.
Why can't I just buy the cooling startup, like the one that makes the chip microfluidics?
Because it's usually private. Microsoft's in-chip cooling was developed with a small Swiss startup that has no public shares. And when promising cooling specialists do get big enough to notice, they tend to get acquired by a large public company — LiquidStack, Boyd's thermal unit, and CoolIT all got bought recently. Often the play is to own the acquirer, not the startup.
Can't they just use the waste heat to make electricity or heat homes?
Heating homes, yes — Finland and a few other places pipe data center heat into district heating systems, and it works well where that infrastructure already exists. Making electricity from it, mostly no. The heat coming off these chips isn't hot enough to spin a steam turbine efficiently; thermodynamics gets in the way. So the realistic win is reusing the warmth, not converting it back into power.
Isn't dumping all that heat into the ocean an environmental problem?
It's the top concern people raise, and it's fair to ask. The counterargument is that the ocean's thermal mass is enormous compared to the air, so a given amount of heat raises its temperature far less — and an underwater center can use less energy to cool itself in the first place. But it's genuinely unsettled, which is part of why nobody's rushing to scale it in Western waters yet.

Keep reading

Byte Bungalow

I do the homework out loud.

If this is the kind of thing you want checked against the documents instead of the hype, the video's above — and there's a new one every week.

Subscribe on YouTube →

By Byte Bungalow. Home power and home tech, checked against the documents instead of the hype. Independent commentary; not affiliated with any manufacturer, utility, or builder named here. Not professional electrical advice.