AIBU to question who pays for AI data centre costs?

[Mamalicious16]

AI data centres impact - guess who's paying the bill?
Us! The infrastructure puts a huge strain on our water supply and national grid - and who's paying for these extra electricity and water costs? The general public- how is that fair?

Obviously. But you're using a computer or smartphone to use this site, so you're using AI yourself. So arguably, of course you should be paying for it. The same as people complained about having to pay the additional cost of paved roads when they came along. Technology develops, there's costs and benefits to everyone.

The environment.

I'm not a technical person so I've installed DuckDuckgo as my web browser. I wasn't asked if I wanted AI on my phone. I'm not sure I can avoid it / install it. If the cost is so huge, I don't want it. And I don't see why AI trillion dollar companies shouldn't fund it themselves/ make people subscribe - we already have such a drain on our resources and bills are rising - sounds like we have to pay in one way or another whether we like it or not

No you are not using AI to simply use a computer or phone to browse a website. Some websites and apps might have unavoidable AI features, but the vast majority of websites and apps used daily by people do not use AI.

AI datacentres are completely different from traditional datacentres that run websites, databases, busines applications etc. AI datacentres require water cooling whereas traditional servers use air cooling. AI datacentres use VASTLY more power (the same as a small town) compared to traditional datacentres. A rack full of traditional servers will run at about 3kW - the same as a kettle. A rack full of GPU-loaded servers for AI workloads runs at 150kW.

AI datacentres are completely different from traditional datacentres that run websites, databases, busines applications etc. AI datacentres require water cooling whereas traditional servers use air cooling. AI datacentres use VASTLY more power (the same as a small town) compared to traditional datacentres. A rack full of traditional servers will run at about 3kW - the same as a kettle. A rack full of GPU-loaded servers for AI workloads runs at 150kW

AI, modern search engines etc, all use for more computing power than computing 20 years ago had. Heating and cooling has always been an issue in datacentres. They are usually designed with hot and cold aisles, to draw in cooler air and push out the hot air in the same direction.

Water cooling isn't that new - we considered it as an option about 20 years ago when installing new servers. The datacentre I currently work in has some water cooling. Obviously water in a datacentre increases some risk - water and computers are generally not a good mix, at least if it goes wrong. Mind you, the time we lost cooling to a whole datacentre was probably the most challenging I've dealt with in my career. Overheating servers shut themselves down, not always in a controlled manner.

You can get the specs for servers online. Thry usually mention their KW usage, and also BTU output. Obviously you are unlikely to know what servers and storage arrays any given supplier is using, but you can get an idea of enterprise level specifications.

Companies are working on technology that is more energy efficient, like low-energy chips, but it will take a few years for it to become widely used.

We all use the data- we're on Mumsnet, for a start - so yes, we consumers should be paying towards it.

Bot not at a cost to the national grid or our water supply. Or our bills. Many of us are struggling already. We should have a choice whether we use AI or not. MN has been going for years, surely it hasn't always used AI. In my view there should be clearer transparency ( eg a banner that tells people AI is used and have an option to opt out. I'd be happy to use a more ( chunky?) MN if I had an option. Is it clear for everyone else? Maybe everyone else is aware that it's using AI and the impact it has and I've been living in a cave.

It would seem logical to me to build them near the coast - use hydroelectric power generators and sea water (to be recycled) for cooling. This would cut both costs and environmental impact, so we’d pay less from every perspective.

Maybe this has already been considered or isn’t feasible.

We all do. All so that we can be dumbed down as much as possible. O brave new world.

You’re being ridiculous. There can’t be an option to use a website with AI and one without just because you don’t want to move with the times.

Do you even realise what it takes to maintain a website, particularly of this size?! You’d need double of everything just to facilitate your whinging of not wanting to use AI.

You‘re also using technology, data centres are not just about AI. Opting out of AI doesn’t mean no part of your usage is serviced through a data centre. I see no reason you wouldn’t also pay.

As far as I'm aware datacentes don't consume water, ie make it not water any more, they use water but don't reduce the overall amount in the world so it's taking it out of circulation that might be an issue

I read on a previous thread that golf courses use many many times more water than data centres, I looked it up and it appears to be the case. Am interesting statistics I thought

It’s a very important question, because the cost of an AI data centre is much bigger than the building and computer equipment. Someone has to pay for the additional electricity generation, transmission lines, substations, water systems, roads, wastewater treatment, etc.
The short answer is: it depends on the country, utility and contract — and sometimes ordinary taxpayers and utility customers end up paying part of it.

There is a lot more I can tell you if you ask me, but the answer is to long to post here.

Ask 'Who pays for all the additional infrastructure costs to support the AI Data Centres, water and electricity supply etc.'

Can the hot water not be used in the local energy network?

@EBearhug I'm well aware, I've been in the industry since the early 2000s. But just look at the power consumption of a rack of say DL380s compared to an Nvidia DGX rack - the numbers are eye-watering.

AI is becoming thoroughly embedded into the process of software development. If you are using software that has been developed or updated in the past few years then you are indirectly using AI (this should be all software you use since you don’t want to be using out of date software due to security concerns).