This data center boom is so reminiscent of the .com period. Suddenly everybody with two computers next to each other had a data center (usually in a closet somewhere, if you were lucky it was locked).
I was working at a startup that was being acquired and as part of due diligence they sent someone to do a "walk through" of our "data center," which was tricky because it was in fact a small coat closet with 2 beige tower PCs stacked up in it.
I actually wish I could round up every computer in my home, stick them in a 19-inch rack, and put that rack someplace where I can't hear it. That would be my "data center". It will be fed with a fat power cable and a nice fiber uplink, and it might even run some "AI" from time to time.
2026 is going to be about going all in on automating software development. This is where the money is currently at as software developers need compute and what could justify all the data centers being built. It is much harder to convert students with limited money or people who don't care about accuracy or the highest end models as much in a lot of other fields. The thing most people even got excited about by the end of this year is obviously Claude Code. If they can solve software development the bubble may take longer than expected to burst.
> I am also deeply concerned about the “speculative” data center market. The “build it and they will come”strategy is a trap. If you are a hyperscaler, you will own your own data centers.
Is this actually true? I thought that hyperscalers keep datacenters at arm's length, using subsidiaries and outsourcing a lot of things.
Hyperscalers use various subsidiaries and shell companies to dodge taxes and keep the debt off their balance sheets so they can keep their AAA ratings, but ultimately the resulting datacenters are still entirely owned and operated in-house. Hyperscales do not and will not use any of these “DC as a service” startups, meaning those startups have to find customers elsewhere; the big question mark is whether enough of those customers exist.
If/when the AI bubble pops and we're left with hundreds of underutilized datacenters, I think this could be good for hosting customers by driving prices down across the board. Maybe it will make self-hosting even more cheaper than public clouds.
The GPUs being used for AI training aren't useful for running games, I highly doubt they could be repurposed for that.
They can run other ML stuff but I don't see how the world could absorb the amount of compute/RAM for these other workflows.
Since the hardware itself depreciates quite fast compared to the buildings, the long lasting physical structure is probably the best bet, being repurposed for more general computing (since all HVAC, electricity, and other expensive infrastructure is already in place).
I think it will also allow later down the board for more competition down the line for the hardware 3-5 years down the line.
But I don't really think that focusing on GPU market is gonna work as I don't really see much reason to cater to GPU intensive workloads unless you are affiliated with AI (for most part, some GPU compute is good but datacenters shouldn't really invest so much in GPU compute)
I feel like I am more optimistic about buying stuff (preferably compute) 3-5 years down the board when things might become cheap in my opinion
But this is still a very long time. On one hand I want Ai bubble to burst asap to lessen the impacts of financial crisis but on the other I really think about the financial crisis and think if there are any ways to mitigate the economy's loss so that it doesn't become 2008 crisis
But reality to me feels as if the chances of Bubble popping up is inevitable, the only question is if we can do anything to lessen the strain on the average person all around the world. I think America will be impacted the most but I wonder how geo-politically this might impact other countries as compared to the 2008 crisis.
I don't really think the american govt. is doing anything to lessen the strain the bubble can cause tho, perhaps even promoting things like stargate/AI bubble itself.
It's just sad for 3-4 years. Lets hope economy gets better and more reasonable than AI hype
Don't these data centers have pretty elaborate cooling setups that use large volumes of water?
So they're sitting on real estate with access to massive amounts of water, electricity, and high bandwidth network connections. Seems like that combination of resources could be useful for a lot of other things beyond just data centers.
Like you could probably run desalination plants, large scale hydroponic farms, semiconductor manufacturing, or chemical processing facilities. Anything that needs the trifecta of heavy power, water infrastructure, and fiber connectivity could slot right in.
> Don't these data centers have pretty elaborate cooling setups that use large volumes of water?
Depending on where, and (more importantly) when you last read about this, there's been some developments. The original book that started this had a unit conversion error, and the reported numbers were off by about 4500x what the true numbers are (author claimed 1000 times more water than an entire city consumption, while in reality it was estimated at ~22% of that usage).
The problem is that we're living in the era of rage reporting, and corrections rarely get the same coverage as the initial shock claim.
On top of this, DCs don't make water "disappear", in the same way farming doesn't make it disappear. It re-enters the cycle via evaporation. (also, on the topic of farming, don't look up how much water it takes to grow nuts or avocados. That's an unpopular topic, apparently)
And thirdly, DCs use evaporative cooling because it's more efficient. They could, if push came to shove, not use that. And they do, when placed in areas without adequate water supply, use regular cooling.
Water usage of the DC itself can vary a lot. If they're in an area where clean water is cheap, then they might use evaporative cooling which probably has the most significant water consumption (by volume and the fact that it's been processed to be safe to drink). In other areas they may use non-potable water or just a closed loop water system where the water usage is pretty negligible. The electricity is going to be the much larger consideration on the larger scale (though still affected by local grid capacity). Also, the capital cost is a very significant part of these systems: there's a pretty big gap in pricing between 'worth building' and 'worth keeping running'.
(I recommend this video by Hank Green on the subject: https://www.youtube.com/watch?v=H_c6MWk7PQc . Water usage of data centers is a complex and quite localized concern, not something that's going to be a constant across every deployment)
Semiconductor manufacturing might make sense here but I also don't think it might not simply because it would require probably a lot of expertise and knowledge and complex machinery with experience in this industry which I assume would be very hard to gather even for these datacenters.
I don't see any reasonable path moving forward for these datacenters for the amount of money that they have invested.
Semiconductor manufacturing needs supply chain a lot more than it needs fast internet. Wafers, fine chemicals, gases, consumable parts. A lot of this comes from petroleum refining, so it helps to be near a lot of refineries, although not enough to be decisive in site selection.
I think the interesting thing would be if something like Euclyd is successful, and the 10 GW become 100 MW, and the giant server halls become 800 cabinets total.
Then the inference mega datacentre investment becomes a losing proposition for those who paid too much for the wrong kind of computers, but AI users still get more compute, lower prices etc.
The overlap between the datacenter needs of AI and those of crypto are hard to miss. If/when the AI bubble pops I'd be willing to bet that we see another crypto surge.
I can't be the only one who's noticed that the volume of crypto stuff being pushed slowed as the AI fervor ramped up?
Many of the crazy things we see happening today come from the same root problem: the wealth of humanity has increasingly been funneled into the hands of a few. These may be mostly companies rather than individuals, but they are a proxy for a relatively small few people.
This glut of finances leaves those few with a problem: what to do with it? And since accumulation of wealth is the ultimate goal, FOMO encourages increasingly foolish decisions (or at least unsustainable ones).
The housing crisis is one of the outcomes. This AI bubble is another. I'm sure if we look around, we can find other examples.
If resources were more evenly distributed, we would see different regions and peoples doing more varied things which would be appropriate for their needs.
Axios hijacks the back button. The TLDR is: we’re building too much speculative capacity and the “if we build it, they will come” assumption will be proven wrong.
It's always a bug. Usually, some issue with an iframe navigating to another html page, which makes the back button go back in the iframe instead of the page itself. This usually happens because of an ad, which makes me wonder why people are still navigating without ad blockers.
They are building too many steam engines. It's a bubble that's going to pop and leave investors empty handed. There are only so many mines to pump water out of, mills to turn, textile machines to crank. And those industries have jumped on a bandwagon that they don't understand, and most will soon mostly revert to good old English laborers. His Majesty's subjects are starting to see the terrible environmental effects of burning so much more coal. As a prudent investor I now recommend a Sell for the stocks of the Boulton & Watt company.
This did pretty much happen with railways during the industrial revolution. Huge amount of money was invested into railway build-outs and most investors lost their money even though the result was useful in the long run (though only a subset of what was built).
> Data centers are becoming political flashpoints, primarily because of their impact on electricity prices.
I'd bet anything that they know people are pissed the market has wildly misallocated capital and are trying to save face by framing it as a cost that directly impacts consumers. In reality, the fear is knowing that our investor class is really too dumb to handle the money we've stuck in a box called "the world's retirement funds" and handed to them.
Hmmm, I don't think this is the average perception of it, though. In general people I've discussed this with are more enraged by the resource usage (whether from a price increase or environmental position) than the capital.
This data center boom is so reminiscent of the .com period. Suddenly everybody with two computers next to each other had a data center (usually in a closet somewhere, if you were lucky it was locked).
I was working at a startup that was being acquired and as part of due diligence they sent someone to do a "walk through" of our "data center," which was tricky because it was in fact a small coat closet with 2 beige tower PCs stacked up in it.
The future of the economy is everyone owning 4 data centers
I actually wish I could round up every computer in my home, stick them in a 19-inch rack, and put that rack someplace where I can't hear it. That would be my "data center". It will be fed with a fat power cable and a nice fiber uplink, and it might even run some "AI" from time to time.
2026 is going to be about going all in on automating software development. This is where the money is currently at as software developers need compute and what could justify all the data centers being built. It is much harder to convert students with limited money or people who don't care about accuracy or the highest end models as much in a lot of other fields. The thing most people even got excited about by the end of this year is obviously Claude Code. If they can solve software development the bubble may take longer than expected to burst.
> I am also deeply concerned about the “speculative” data center market. The “build it and they will come”strategy is a trap. If you are a hyperscaler, you will own your own data centers.
Is this actually true? I thought that hyperscalers keep datacenters at arm's length, using subsidiaries and outsourcing a lot of things.
Hyperscalers use various subsidiaries and shell companies to dodge taxes and keep the debt off their balance sheets so they can keep their AAA ratings, but ultimately the resulting datacenters are still entirely owned and operated in-house. Hyperscales do not and will not use any of these “DC as a service” startups, meaning those startups have to find customers elsewhere; the big question mark is whether enough of those customers exist.
> We will stay largely under the radar
He says this and then publishes the letter on LinkedIn? What's the play?
I am hoping things revert to the mean and by extension, the consumer chip prices revert.
with investors and hype, things regress to the extreme
Absurd. Datacenter requirements are only going up, and will continue to go up.
If/when the AI bubble pops and we're left with hundreds of underutilized datacenters, I think this could be good for hosting customers by driving prices down across the board. Maybe it will make self-hosting even more cheaper than public clouds.
Aren't these datacenters biased majorly towards GPU capacity? Maybe we'll see a resurgence of cloud game streaming initiatives.
The GPUs being used for AI training aren't useful for running games, I highly doubt they could be repurposed for that.
They can run other ML stuff but I don't see how the world could absorb the amount of compute/RAM for these other workflows.
Since the hardware itself depreciates quite fast compared to the buildings, the long lasting physical structure is probably the best bet, being repurposed for more general computing (since all HVAC, electricity, and other expensive infrastructure is already in place).
I think it will also allow later down the board for more competition down the line for the hardware 3-5 years down the line.
But I don't really think that focusing on GPU market is gonna work as I don't really see much reason to cater to GPU intensive workloads unless you are affiliated with AI (for most part, some GPU compute is good but datacenters shouldn't really invest so much in GPU compute)
I feel like I am more optimistic about buying stuff (preferably compute) 3-5 years down the board when things might become cheap in my opinion
But this is still a very long time. On one hand I want Ai bubble to burst asap to lessen the impacts of financial crisis but on the other I really think about the financial crisis and think if there are any ways to mitigate the economy's loss so that it doesn't become 2008 crisis
But reality to me feels as if the chances of Bubble popping up is inevitable, the only question is if we can do anything to lessen the strain on the average person all around the world. I think America will be impacted the most but I wonder how geo-politically this might impact other countries as compared to the 2008 crisis.
I don't really think the american govt. is doing anything to lessen the strain the bubble can cause tho, perhaps even promoting things like stargate/AI bubble itself.
It's just sad for 3-4 years. Lets hope economy gets better and more reasonable than AI hype
Don't these data centers have pretty elaborate cooling setups that use large volumes of water?
So they're sitting on real estate with access to massive amounts of water, electricity, and high bandwidth network connections. Seems like that combination of resources could be useful for a lot of other things beyond just data centers.
Like you could probably run desalination plants, large scale hydroponic farms, semiconductor manufacturing, or chemical processing facilities. Anything that needs the trifecta of heavy power, water infrastructure, and fiber connectivity could slot right in.
> Don't these data centers have pretty elaborate cooling setups that use large volumes of water?
Depending on where, and (more importantly) when you last read about this, there's been some developments. The original book that started this had a unit conversion error, and the reported numbers were off by about 4500x what the true numbers are (author claimed 1000 times more water than an entire city consumption, while in reality it was estimated at ~22% of that usage).
The problem is that we're living in the era of rage reporting, and corrections rarely get the same coverage as the initial shock claim.
On top of this, DCs don't make water "disappear", in the same way farming doesn't make it disappear. It re-enters the cycle via evaporation. (also, on the topic of farming, don't look up how much water it takes to grow nuts or avocados. That's an unpopular topic, apparently)
And thirdly, DCs use evaporative cooling because it's more efficient. They could, if push came to shove, not use that. And they do, when placed in areas without adequate water supply, use regular cooling.
Water usage of the DC itself can vary a lot. If they're in an area where clean water is cheap, then they might use evaporative cooling which probably has the most significant water consumption (by volume and the fact that it's been processed to be safe to drink). In other areas they may use non-potable water or just a closed loop water system where the water usage is pretty negligible. The electricity is going to be the much larger consideration on the larger scale (though still affected by local grid capacity). Also, the capital cost is a very significant part of these systems: there's a pretty big gap in pricing between 'worth building' and 'worth keeping running'.
(I recommend this video by Hank Green on the subject: https://www.youtube.com/watch?v=H_c6MWk7PQc . Water usage of data centers is a complex and quite localized concern, not something that's going to be a constant across every deployment)
Semiconductor manufacturing might make sense here but I also don't think it might not simply because it would require probably a lot of expertise and knowledge and complex machinery with experience in this industry which I assume would be very hard to gather even for these datacenters.
I don't see any reasonable path moving forward for these datacenters for the amount of money that they have invested.
Semiconductor manufacturing needs supply chain a lot more than it needs fast internet. Wafers, fine chemicals, gases, consumable parts. A lot of this comes from petroleum refining, so it helps to be near a lot of refineries, although not enough to be decisive in site selection.
I think the interesting thing would be if something like Euclyd is successful, and the 10 GW become 100 MW, and the giant server halls become 800 cabinets total.
Then the inference mega datacentre investment becomes a losing proposition for those who paid too much for the wrong kind of computers, but AI users still get more compute, lower prices etc.
The overlap between the datacenter needs of AI and those of crypto are hard to miss. If/when the AI bubble pops I'd be willing to bet that we see another crypto surge.
I can't be the only one who's noticed that the volume of crypto stuff being pushed slowed as the AI fervor ramped up?
I'm sure this will be a nice consolation as the meaningful part of the economy collapses
I was more thinking these data centers will become homeless shelters.
I feel the same way... Posted this yesterday:
After "AI": Anticipating a post-LLM science and technology revolution (evalapply.org)
https://news.ycombinator.com/item?id=46419416https://www.evalapply.org/posts/after-ai/index.html
> I, for one, welcome the coming age of the post-LLM-datacenter-overinvestment-bust-fueled backyard GPU supercomputer revolution.
The best indication that this isn't a bubble is all the people who would jump at the chance to buy more capacity if the price drops even a little.
Only if they were buying GPU capacity.
Many of the crazy things we see happening today come from the same root problem: the wealth of humanity has increasingly been funneled into the hands of a few. These may be mostly companies rather than individuals, but they are a proxy for a relatively small few people.
This glut of finances leaves those few with a problem: what to do with it? And since accumulation of wealth is the ultimate goal, FOMO encourages increasingly foolish decisions (or at least unsustainable ones).
The housing crisis is one of the outcomes. This AI bubble is another. I'm sure if we look around, we can find other examples.
If resources were more evenly distributed, we would see different regions and peoples doing more varied things which would be appropriate for their needs.
We need a concentration of money in order to make progress.
But yes, we should also make sure that everybody can benefit from it.
>We need a concentration of money in order to make progress.
...in the concentration of money.
"The handful of benevolent god-kings will be nice to people" is not a model for a decent future.
You mean we need a stronger government?
Axios hijacks the back button. The TLDR is: we’re building too much speculative capacity and the “if we build it, they will come” assumption will be proven wrong.
Such bad behavior on the back button - it’s such a shabby trick, and makes me think less of any site that pulls that.
I always think about the developers who were asked to implement something like this
It's always a bug. Usually, some issue with an iframe navigating to another html page, which makes the back button go back in the iframe instead of the page itself. This usually happens because of an ad, which makes me wonder why people are still navigating without ad blockers.
Happens to me even with an ad blocker.
Sorry, what do you mean by this? Back button seems to work normally for me.
Yeah annoying. Need to find an extension to fix that.
They are building too many steam engines. It's a bubble that's going to pop and leave investors empty handed. There are only so many mines to pump water out of, mills to turn, textile machines to crank. And those industries have jumped on a bandwagon that they don't understand, and most will soon mostly revert to good old English laborers. His Majesty's subjects are starting to see the terrible environmental effects of burning so much more coal. As a prudent investor I now recommend a Sell for the stocks of the Boulton & Watt company.
This did pretty much happen with railways during the industrial revolution. Huge amount of money was invested into railway build-outs and most investors lost their money even though the result was useful in the long run (though only a subset of what was built).
It also happened with fiber during the dotcom boom. Years later, that came useful when everyone wanted YouTube in their pocket.
> They are building too many steam engines. It's a bubble that's going to pop and leave investors empty handed.
What a great analogy. People building too many steam engines was one of the main factors that lead to the first world war.
Well over a century after the commercialization of the Newcomen / Watt steam engines.
Sounds interesting. Do you have a source for that?
> Data centers are becoming political flashpoints, primarily because of their impact on electricity prices.
I'd bet anything that they know people are pissed the market has wildly misallocated capital and are trying to save face by framing it as a cost that directly impacts consumers. In reality, the fear is knowing that our investor class is really too dumb to handle the money we've stuck in a box called "the world's retirement funds" and handed to them.
"our investor class is really too dumb to handle the money we've stuck in a box called "the world's retirement funds" and handed to them."
all it took was a chatbot that says "you're a genius"
Hmmm, I don't think this is the average perception of it, though. In general people I've discussed this with are more enraged by the resource usage (whether from a price increase or environmental position) than the capital.