Nope. You probably can't cash in by turning your office or farm into a datacenter Sorry to be the bearer of bad news, but if you own real estate and think you can cash in by using it to host an AI datacenter, you're probably wrong. That's the opinion of Billy Lee Kok Chi, chairman of Malaysian datacenter builder and operator CSF Advisers, as expressed on Wednesday during the closing plenary of the Asia …
One thing I remember from the peculiar units† rote learnt during my long ago £/s/d schooldays is that there were 640 acres in a square mile. (2.56 square km or 1.6 km × 1.6 km)
The idea that this AI madness is going to consume 365.25x24 GWh‡ of energy every year and cover a square mile of probably otherwise quite productive land is insane and all just so some corporate, utility or government instrumentality can have an AI agent which will give you the same pre AI runaround.
† rod (aka pole or perch), chain, furlong; perch, rood, acre but mercifully no hides.
‡ I guess enough energy to heat 75 Mt of water 100°C.
Starting with Trump's golf courses.
Carving out an exception for the Trump International Links near Aberdeen, which was/is (in places) a fragile and rare coastal dune system that they drove a links through the middle of (and then whinged and bitched about an offshore wind farm "ruining the view"). So restore that to it's proper natural habitat and turn the club house into an environmental education centre.
Rest of them, yeah fine. Nothing of value there.
Square mile of data centre = lots of roof space to lay out a "high-level scenic[1] links". With water traps (cooling ponds) and sand bunkers (ready to dump onto a battery fire).
Your scores will be used to feed into the random number generators. Your distinctiveness will be used to train the models. Please do not read the final page of the member's agreement detailing the use of Cyber Conversion in case of spontaneous AGI.
[1] You can't see that godawful ugly DC, unlike everyone else in the vicinity, because you are on top of it!
High liklihood of failure in England - Golf Courses are frequently used as illegal dumping sites for industrial waste during their construction to build the hazards - thereby creating another hazard later on (groundwater leaching is turning out to be a BIG issue)
"....there were 640 acres in a square mile"
Also known as a "section" in Canada particularly in the prairies. So if a Saskatchewanian farmer tells you he has a 10 section farm, he's not using a 1950 Massey Fergusson to plant wheat :)
"but mercifully no hides"
They were variable depending on the quality of the land, at least initially when a hide was the land needed to support a family and could be ploughed by a team of 8 oxen over the ploughing season.
We have an awful lot of empty land in northern Canada, we just need to build a nuke plant to power it Then instead of the cooling just being an energy sink, the waste heat can be used for district heating, since heating is needed for most of the year, most of the waste heat won't be wasted
Alot of us knew this. Over on the selfhosting subreddit, you'll get the odd person ask if they can make money renting out their selfhosting and some on the datahoarder have mentioned it. Answer is always no. Because you then need to hire lawyers, put in redundancy, proper backups and proper user contracts.
Use AI to tell you how many feet above the current sea level you need to keep the new AI data-center not flooded ... looking at the current climate change predictions it looks like being at least 35 feet above the sea level would work, so I guess no chance to a setup in the Netherlands?
The premise this article takes for granted is fictitious. We need a bunch more data centres because...? LLMs are limited application AI and we're using them for at least half the things they apply to now. More efficient models are being built now. Inference can be done on commodity hardware. What do we need a bunch more data centres for?
I suspect Moores law + more efficient models will lead to it being more and more possible to simply run your own LLM locally. Which some people already are doing. Whether you bother to run it locally is beside the point. That will decouple it from big tech dominance and make hosting it in a data center less energy intensive.
By 'ignore' I mean refuse to pay for it. There is no business case for AI given the cost of the data centres. They can add it to stuff against our will, but all they are doing is increasing their overheads with no additional return. Forcing a product on people that they do not want and do not trust is a really good way of losing money. How well are those Clippy AI PCs selling? How popular is W11?
Social media is popular because we do not hand over cash for it. Most people are happy to hand over data for it, and data pays for it. Data will never pay for AI data centres. They cost too much.
Data centres for AI is perhaps the biggest money pit of them all. Only governments will build stuff like this as part of nationalist pissing competitions, because they can filch public money to do it.
Tulips, the South Sea Bubble, railway mania. Next, AI. Stock up on popcorn for the inevitable.
Supercomputers are needed because weather forecasts (right pondians take that probably far too seriously) and the stuff particle physicists get really excited about.
AI is mostly a solution looking for a problem. I say mostly because applications such as cancer hunting have real purpose, but helping me write my next slide deck is not. There are things people in the walk if life need to learn; you wouldn’t go to Dragons Den / Shark Tank without very careful honing if your pitch, and AI doesn’t give you that.
I've been away from the business for years now and so am surprised by "AI datacenters that require 230 kW of available energy per rack.", even allowing for AC, UPS and other ancillary nonsense, this seems somewhat excessive. Would someone care to enlighten me?
I'm guessing this is overall power demand for the DC, including cooling, losses in the UPS, redundant supply loops, etc - not the actual per-rack supply. Overall budget per rack.
The biggest, baddest AI cards at the moment (H200, B200 type stuff) are 1kW.
The 6U Dell PowerEdge XE9680 can host 8 such GPUs (e.g. H200). So that's 1.3kW per U just for GPU, nevermind CPU, RAM, storage and networking.
The 2U Dell PowerEdge XE9640 can host 4x 700W GPUs (H100), so that's 1.4kW per U for the GPU.
Of course, realistically all those ancillaries are unlikely to take the total past 1.5kW/U - the GPUs are the stars of the show.
But if we take a Joseph Bazelgette "only-doing-this-once" approach and assume we need (or may need in the future) 3kW/U once we include CPU, RAM, losses at the PSU, etc, etc then that still only gets us to 126kW per rack (assuming standard 42U racks, which a hyperscaler doing a custom AI-DC build might not be...). Which is just about halfway there. This potentially makes sense if you have two redundant power supplies to the building (whether that's distinct grid connections or grid + local generation), each of which must be capable of keeping the lights on. If you need ~120kW per rack, then you budget ~240kW per rack of supply to the building, but you'll never pull that. You'll be pulling half that, and if one side falls over, the other side can manage alone.
Also, the guy is in the business of building out new DCs for clients. He doesn't want people converting existing buildings. So his is an informed - but biased - opinion.
Nvidia mentions a max of 4 x DGX B200 in a 52U Rack.
Each 14.3kW max.
with higher than normal cooling and power delivery required.
They'd need something very specialised for 100kW, you'd think.
El Reg has had articles before about datacenters like Colovere advertising 200 kW potential per rack using liquid cooling. Not equipment actually drawing that power though.
Go figure.
I also agree with others that efficiency is going to improve, you'll be able to do a lot more with a lot less.
As for buildings not being able to take the weight and warehouses not having enough power - I give you AQL's Salem Chapel built in 1791 - which houses a DC amongst other things.
City Centre costs? How many boarded up stores / closed shopping centres/malls are there? Ground floor for racks, 2nd floor for accommodation with free heating ;-).
What else is it going to be used for when everyone is shopping on line :-(.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
