People on this forum should definitely be concerned about the techlash coming for data centers, I doubt many here would enjoy a future where compute follows the same trend as housing prices because it’s illegal to build more of either.
1ft at 2075 assumes we curb emissions somewhat:
https://www.climate.gov/media/14136
Article:
https://www.climate.gov/news-features/understanding-climate/...
Datasource:
https://earth.gov/sealevel/us/internal_resources/756/noaa-no...
After digging into it a bit to find a better source for you, it turns out that my number was wrong anyway. Turns out the sea level rise for the contiguous US is expected to be quite a bit higher than the global average. I had no idea!
That said, I don't think they assume our emissions trend from the last 50 years will continue unabated.
100 local people to maintain the data center while it replaces 1 million people with the AIs running inside
So, no, because said human capital is holding shorter end of the stick and will be worst off.
In my opinion, Compute-related data centers are a good product tho. Offering up some gpu services might be good but honestly I will tell you what happened (similar to another comment I wrote)
AI gave these data centers companies tons of money (or they borrowed) and then they brought gpus from nvidias and became gpu-centric (also AI centric) to jump even more on the hype
these are bad The core offering of datacenters to me feels like it should be normal form of compute (CPU,ram,storage,as an example yabs performance of the whole server) and not "just what gpu does it have"
Offering up some gpu on the side is perfectly reasonable to me if need be perhaps where the workloads might need some gpu but overall compute oriented datacenters seem nice.
Hetzner is a fan favourite now (which I deeply respect) and for good measure and I feel like their modelling is pretty understandable, They offer GPU's too iirc but you can just tell from their website that they love compute too
Honestly the same is true for most Independent cloud providers. The only places where we see a complete saturation of AI centric data centers is probably the American trifecta (Google,azure and amazon) and Of course nvidia,oracle etc.
Compute oriented small-to-indie data centers/racks are definitely pleasant although that market has raced to the bottom, but only because let's be really honest, The real incentives for building softwares happens when VSCode forks make billions so people (techies atleast) usually question such path and non-techies just don't know how to sell/compete in the online marketplaces usually.
Suddenly adding 50gw of power demand in a state is going to drive up costs significantly.
My point is, power generation infra costs money, and currently its being paid for by someoneelse.
During the microcomputer revolution, hackers scoffed at people who used terminals to access time sharing systems. You don't own it, you don't control it, you're just a cog in the machine. Now, "hackers" are rushing to run everything on hardware owned and operated by companies with wealth and power that make the old IBM look like a kid's lemonade stand by comparison.
"It is difficult to get a man to understand something, when his salary depends on his not understanding it." - Upton Sinclair
I feel like people have problem with AI oriented datacenters (which is becoming the majority of datacenters considering that datacenters make an shit ton of money selling AI aka shovels during gold rush)
Another thing is that these datacenters have very high levels of compute directly linked to the consumer of an application
As an example, you have a simple app, some message gets pushed by customer or database query or simple usage, its all good, at a datacenter level its power costs are miniscule
Now compare that to datacenters which have gpu's so they have applications like chatgpt (let's imagine) running on them, now these AI services are used by people themselves.
Now instead of simple applications and executions, Perhaps a trillion parameters models are running now. These are beyond computationally expensive even if we compare them to normal applications
Now I just searched and google's gemini runs 1.5 BILLION such prompts per day and chatgpt runs 2.5 BILLION prompts per day
Now, these prompts, they aren't stable all around the day, I have heard these to be very varying and when power consumption varies, it really impacts the performance of the grid itself
Another aspect is the sheer size, One would imagine that AI Bubble might give them more money and it does but the energy costs seems to me to be so high and perhaps also the fact that AI bubble gives these companies tons of free money which they "invest" aka buy/(lease?) year govt. contracts a lot of electricity
The govt. can only build so much capacity for these electricity and they (lobbying? and many other efforts) when get sold to datacenters really strains the electricity which thus increases the rates of electricity (and in a similar fashion perhaps water too) for the average american.
TLDR the way I read it: compute is cheap. There are always gonna be refurbished old compute which is gonna be too "old" (3-5 years because of deprication but that hardware is a beast) for these guys to use.
Nothing stops a simple guy who loves tech to open a mini datacenter perhaps :)
Who knows what might happen and I was extremely pessimistic about the datacenters not for these reasons but rather that ram prices were rising and I was worried that the whole industry might increase compute prices too but it seems that asus is opening up their ram production for consumers so starting out datacenters is possible
let's see what happens though. And I was worried a bit same as you but I feel like compute prices themselves are pretty chill/can remain chill. I understand the worries tho so looking forward to a discussion about it.
And so on
I'm not an engineer, but it seems hard to imagine that a lack of data center capacity won't have an effect on prices for cloud compute, which will have downstream impact on what workstations have access to (especially since more and more programmers are becoming reliant on coding LLMs).