Random Renewables Thread

Article on RTE today:

http://www.rte.ie/news/business/2024/0807/1463834-eirgrid-warned-of-possible-mass-exodus-of-data-centres/

I have to say, I'm quite annoyed at Eirgrids position here. They basically seem to take the stance that data centres are good and we should have more, and therefore we need more immediate generation with no regard for what's fueling it (i.e. more fossils)

Now I would generally agree that data centres are good economically, both directly through job creation and indirectly by encouraging tech companies to set up here

However, there really needs to be some serious cost-benefit analysis done here

Over the past couple of years the data center expansion, driven largely by AI has been absolutely outrageous. Microsoft alone spent around $19 billion dollars in Q2 2024 on capital expenditure

The energy consumption of all this buildout is nothing short of horrifying, and if you cast your mind back you may remember the rough estimate that 30% of the activity in data centers is non-beneficial automation (spammers, scammers, malware, etc)

There's a serious problem with the value proposition for all this build up activity. No-one seems to be able to come up with some useful business case for AI which is truly world changing. A lot of what the tech bros have said is either easier to do with simple automation, or not worth the cost of development

Apple in particular seem to be treating AI as a useful feature or tool rather than a product in its own right

In this context, Eirgrid should be pushing back on these data center applications and asking whether any of this is sustainable development. Will the cloud providers be scaling back plans in a year or two, will the mass exodus happen anyway as cloud demand stagnates?

In that case, will the Irish rate payer be stuck footing the bill for expensive generation contracts that aren't needed anymore. Will the short term need for generation tie us into more fossil fuel dependence by pulling resources away from renewables?

There's a lot of serious questions need to be asked to Eirgrid IMO, particularly about who exactly they're trying to represent here

imo, the AI workload is going to drop like a stone in a tech stock crash within a year or so (already starting!) and then gradually rebuild when useful use cases become apparent - think a dot.com crash to today timeframe though. But in the meantime, I don’t mind it being used as a Trojan horse to expedite the frankly painfully slow process of connecting clean energy sources. We need to target over-supply of renewables anyway to combat intermittency, may as well spook the political class now to get things moving!

Yes, we really need more intermittent and loss making renewables right now, which is stealing our next PSO levy refund to our utility bills to the tune of 800 wholesale Kwh. That's a quarter of average domestic annual consumption. The PSO levy is set to kick in positive again as wholesale prices return to pre panic levels. Build another Wogawatts of them windmills for the spam farms and misinformation centres, sure we won't mind forking out another €100 levy or so to subsidise these pointless power suckers.

https://www.irishtimes.com/business/2023/08/01/electricity-consumers-will-miss-out-on-pso-levy-refund-next-year/

That's a possibility, however they seemed to be very specifically talking about how zero carbon wasn't an achievable goal for a data center

Okay, I'll concede a zero carbon data center is going to be vey difficult to build. I would expect this kind of talk from a lobbyist, not from Ireland's grid operator

I think the main purpose was to say that Ireland's future generation plans aren't going to support a large scale build up of data centers. My question to that is whether it's truly economically beneficial for that build up to occur here

They also seemed concerned about uncertainty around future generation, which seems fair. There was also some comments about a project hierarchy and what to prioritise, which is also something you'd want to have straight

It's possible they're using data centers as an example to push things along, however they seem to have their own priorities mixed up.

We don't have a national shortage of computing power, whereas we do have national shortages of housing, EV charging, renewable energy as well as a host of old housing stock which is moving towards heat pumps and EVs and may require additional grid capacity

If they'd led with that I suspect there'd be a lot more people on their side. However the cynic in me wonders if Eirgrid make more money from data centers and are trying to shore up their own revenue stream

As simple as…

Solar has a capacity factor in Ireland of 11%, onshore wind is about 28%, Offshore wind should be about 48%.

It might be 'as simple as', if data centres didn't need power 24/7/365 and if you could bridge the yawning chasm by just adding more renewables, but you can't and batteries are nowhere near cheap enough to bridge the gap, despite ridiculous claims to the contrary.

With the obscene amount of money being spent on those data centers they could easily afford some battery capacity for load balancing

At $100 per kWh, Microsoft spent enough money to buy 190GWh of battery capacity in the last 3 months (they spend $19 billion on Capex alone)

That's close to 2.5 million EVs, or all of Tesla's production for 2022 and most of 2023

How many servers will that power? Let's try to figure that out

An Oracle X9-2L server consumes around 600W for a mid range config on "average" load with 100% utilisation according to their power calculator

I know Microsoft don't use Oracle servers but the power consumption should roughly be similar per the same amount of computing power

Assuming you need to spend that amount of power again in cooling, that's a nice 1.2kW per server. I've no idea how cooling scales, but I know of one data center in Chicago that overheated in -30C when the cooling system froze (yep, slight design flaw). So it's fair to say they need a lot of cooling

Our 190GWh will power very appropriately 6.6 million servers for 24 hours, or putting it another way; 13.2 million Intel Xeon processors (2 CPUs per server)

So let's take another direction, if I split $19 billion between CPU cores and batteries, what percentage would I have to spend on batteries to get through say 12 hours?

Well at $10k per Xeon core, $19 billion buys you 1.9 million cores. Doing some reverse engineering, roughly 7% of your Capex would be going on batteries

This is of course assuming 100% of the Capex originally went into CPU cores, which it didn't. However if less went on computing power then you need less batteries, thus lowering the percentage

My point to all of this is that data centres could fairly easily buy up sufficient batteries to cover any peak usage if they were bothered to do so

Yeah I vaguely recall they use batteries that were designed for telecoms applications. They're designed to fit into a cabinet quite easily, and are pretty well orovien at this point

I find it ironic that there's a lot of servers rack style batteries available now, and one would assume the average data center has a lot of server cabinets

So it really shouldn't be a big deal for them to start putting in lithium batteries

Or they could put container sized batteries outside if they're nervous about putting the batteries in the server halls

I'd fully believe the story about the wind turbine. As I recall there were a bunch of data centers that weren't grid connected and used locally generated power from gas turbines

To be clear, I don't have anything against more data centers being built in principle. They do have some economic benefit, but I feel the benefit is quite oversold relative to the environmental costs.

The current building rush is also just a shocking waste of resources IMO. I would be quite happy to see the tech companies being forced into some "exit tax" where they needed to pay out for the infrastructure they had built for them if they try to roll back on their targets

Big construction projects in this country take almost a working lifetime, so its hard to get corporate expertise to commit to that level of project management, when it will see them on a pension before its finished.

The electrification of my commuter line was approved in outline when I was only halfway through my working life. I'm retired 4 years now and not a cable in site. I won a £20 bet with a work colleague (yes, ir£s) that neither of us would work in our new university campus, so long would it take to plan and build. That bet is unpaid, as he died before it opened, and I retired the year before it opened.

It takes us 5 years just to think about overdue projects, another five to set up committees to commission reports, another 5 to read them etc.. Serbia has recently went from inception to operation of hundreds of kms of 200kph high speed intercity rail in 5 years by simply contracting the job out to Chinese expertise. It also came in at a small fraction of the cost. We need a global view on projects like this. There has to be a way access this outside skill without having to apologise and grovel to the US. We also need a fast track system to allow projects get to the shovel stage ASAP. Co.Councils should be totally removed from top tier infrastructure projects, Judicial reviews likewise.

We can't even sort out the mica houses' redress, despite the costs rising to well above the affected homes' full normal market value. Just offer owners that market value to buy a house elsewhere, then take and sell off their crumbling remnants instead of trying to fix them up. Instead, we have meetings and committees and objections and reviews and stagnation.

I've watched this kind of Gombeen approach for an entire lifetime, from being a small child watching our parents and community excitedly announce a new town development fundraising committee to supply us kids with our own swimming pool. It never happened to this day, though decades of fundraising went unaccounted for. There's money to be made in delaying development, at all levels, that's one of the biggest problems.

Weighs about the same as my car but I'm only packing 80kwh 😓

To be fair to the battery in your car, it has a car attached to it

That's gonna add some weight 😉

Also I'm going to guess the static battery isn't designed to withstand an impact at 120km/h without cracking

Or maybe it is, in which case I'll be quite impressed

Perhaps we'll see those battery cabinets mandated into new residential developments, one for every block say, ten to twenty units, with each block an off grid cell, communally feeding the battery with their panels. If you read this below you can see how this might make sense.

https://oilprice.com/Latest-Energy-News/World-News/Negative-Power-Prices-Hit-Europe-as-Renewable-Energy-Floods-the-Grid.html

Sort of makes this deal look like a bargain 😂