Why you should Get to grips with data centre politics

The backlash against data centres is only just reaching the UK but it would be a big mistake to mistake this for ‘just more Nimbyism’. Unless you understand what is different about this backlash you will miss one of the most significant things happening in politics right now.

Last week we did a Daily Briefing on the politics of data centres. I have had correspondence asking if we could expand on this for people so let me do that here. I realise that sometimes I (and indeed all of us at Common Weal) assume that you get to spend all day reading research papers and policy analysis like we do, and therefore wrongly assume you know a lot of this stuff.

So let me take you through a super-fast spin around the wider politics of data centres. I'll start with (a little) computer geek stuff, run through what a data centre is (and isn't), explain the economic context of AI, explain the economic context of 'compute', place that in the modern economy, pull that together to explain the dynamic which is causing the data centre backlash and then show you what that actually looks like and what it could mean.

Most people seem to believe that AI is a form of 'thinking', but that is a mistake. It is much more accurately described as a pattern predictor. All it does is look at a sequence of 'somethings', then looks at lots and lots and lots of other sequences that look similar and then uses them to guess the probability of the next item on the original sequence.

You best know this as 'words' or 'pixels'. You ask it to summarise a question, it guesses the most likely answer to that question based on everything ever written, one word or phrase at a time. Alternatively you ask it to produce a picture and it starts guessing how it should look from your prompt, pixel by pixel.

Practice it enough times, telling the machine what results are most 'sensible looking', and it learns how to produce sensible-looking answers. And a lot of the time they are full-on sensible answers. But (here our problems start), sometimes they are not, or some smaller part of a bigger answer isn't sensible, or the argument is right but one of the numbers is wrong. Because it has only been trained to look sensible, not to understand what it is saying.

And that's what data centres do. This stuff is processor-intensive so it doesn't happen in your computer but over the internet and into one of these big warehouses. This is not data like you think of data. It isn't stored information about you or anything else. It is just a big computer. There are other data centres which store data – mainly the ones which 'store the internet' and those are totally different.

So when you hear 'new data centre', what someone in Big Tech hears is 'compute' (used as a noun), because 'compute' (used as a verb) is what these warehouses do. Tech views a 'server' (storing and 'serving' you the data that is the internet) as a totally different thing. It's just that, for you, both look the same from the outside. Why, therefore, are we all so angry? We weren't attacking server warehouses.

It starts with the economic context of AI. If it wasn't for the when and what of AI, it would probably wouldn't be kicking off like it is. But the when is 'at the lowest point for trust in the economy and in the rich' and what is 'rich people don't want to pay you ever again'.

One theorist suggests you should think of AI in the same context as immigration. Rightly or wrongly, both are seen as someone else wanting to take your job away specifically so you and people like you will never be paid the same again. With AI that is for certain true.

Will AI really take people's jobs at the rate currently believed? Hmmm, it's complicated. I am not a full-on AI cynic, but I am a sceptic. First of all, the evidence of cost saving isn't great so far. But then evidence of performance isn't great either. There aren't yet the use values in replacing humans as a source of language.

But that last point is crucial; remember I wrote that AI is just a pattern predictor? This is important. Language is a purely human pattern and we understand it best, but mathematics, gene sequences, molecular biology, engineering, chip design? These are all different 'language' sequence systems and this is where I think the real benefits come because they're not human languages.

Coding, accountancy, engineering, synthetic biology, micro-engineering, vaccine design – these I expect to be positively impacted in ways that could change everything. But making chatbots less useless? This I am less sure turns the world upside down.

But don't kid on this isn't already a reality which is raging. In cinema, in computer games, in computer programming, in music production, already AI tools are producing content that would have taken a lot of human labour – and people absolutely hate it. There are fights everywhere. The economic context of AI is that it will take jobs and replace them with wealth concentration. I'm just not sure it will happen in the sectors some people think it will. (Bad art is already ubiquitous.)

So this would be incendiary anyway. This could be like the first industrial revolution, or it could be like the enclosures. The former really did transform the world and in large part for the better, where the latter just altered who owned or had access to land. There was no great improvement from the enclosures like from the industrial revolution. Which this looks more like is something we'll find out.

But what to remember is that both of them at the time utterly immiserated vast swathes of the population. Those who celebrate the industrial revolution now never mention the three or four decades worth of people who starved to death as the disruption shook its way through the economy. And we have democracy now.

What this most looks like in the West is the offshoring of the globalisation era and that has basically broken society and opened the door to a furious far right. Multiply that by, well, a lot and there is the AI risk to our society, politically-speaking.

“Data centres are oligarchy in aluminium and steel”

But that isn't really the right economic context either, because it isn't the data centres coming for your job, it's your boss. So far there isn't that much supportable evidence that anger at data centres stems from specific anger about specific job losses. It is a different economic context we have to consider.

This one is best understood by listening to the noise. Who talks about data centres and what does it sound like? Well, it sounds like arrogant rich people talking among themselves about how to get much richer at the expense of ordinary people. Data centres are oligarchy in aluminium and steel.

It is impossible not to see this in the context of what used to be called 'class war'. Why do I say that though? There are also very solid Nimby reasons to oppose data centres. They put an enormous load on local infrastructure, requiring even more invasive infrastructure to power them and greatly increase the risks of water shortages. Plus they create very few jobs. It's all lose-lose.

So why am I arguing for a class war frame rather than just a Nimby one? Because no-one is writing panicked analysis pieces in US political bubbles about 'the worrying rise in furious anti-battery park protests'. Make no mistake, there are protests against other big infrastructure, but for two reasons it is seen differently in politcs.

The first is simply that the narrative fury isn't there – battery plants are ugly but they don't take your jobs... Batteries aren't 'hated', data centres are. And that is also because the politicians are yet again on the oligarch side. This I need to explain.

Fully a third of the US stock markets are now invested in Big Tech and that is only because of the data centre hype. The real economics of data centres is basically just 'subprime' – don't look too closely at the financial sums, just count the total. The total looks like a multi-trillion dollar boost to the economy. But then so did CDOs made up of sub-prime mortgages.

The simple facts are that no AI company is anywhere near having a revenue base that can even begin to justify this scale of speculation. In fact none of them have a business model for explaining how that kind of revenue is ever going to be secured. No-one really knows if AI will work like they claim and no-one really knows how much you can charge for it.

Well, you do if you use a Chinese model because they are open source and therefore free. They're also much more efficient so you don't need the same 'compute'. So what is the US playing at? The usual. Data centres are equal parts hype and a magical money tree like subprime. They work as easy money investments that make you rich – so long as you don't look closely at what you're investing in.

Unfortunately, to execute their 'let's get so rich no-one can count it' plans they need to take AI companies public. That is how to bring in the really big investment, and so three of the biggest players are 'going public' this year in what is known as an IPO (initial public offering). Problem is that to do that they are obliged to open their books to investors, and investors don't love what they're seeing.

It would take Anthropic much more than three centuries to repay the valuation of the company from profits. In fact there is strong evidence that they all know this is going to crash, because they are slightly inexplicably boosting their initial valuations in a way certain to result in a rapid collapse in share values.

Best anyone can work out, the oligarchs plan to take your pension money and money from what in the US are known as 'retail investors' (unlike the UK, lots of people in the US are encouraged to directly own shares). It looks like Joe Public Investor is going to be fleeced out of this (they’ll spend the capital and it is not clear investors will get the value back in the near future).

Meanwhile data centres are being quietly shelved everywhere because of this non-adding-up arithmetic. The hype is reaching fever pitch and deflating, at the same time. Which takes us back to the protests.

I doubt the public has looked at the economics in this way but they seem to have absorbed the jist of it. And what it looks like to them is exactly the off-shoring of US jobs to China, but rather than China it is a bank of computers, and rather than it being at the other side of the world, it is being put on your doorstep.

And the people who are putting it on your doorstep are openly explaining how much of your water they want and how much energy they will take from you. And then they're audibly bragging about all of it and how it is going to make them rich by taking your jobs away. And just as a cherry on the top, there aren't even any jobs in data centres. They sit there taking your jobs, your water, your electricity and your views and you get poverty.

This has absolutely freaked the politicians out in the US. They were busy celebrating the numbers and were ignoring whose backs the numbers were being placed on. Then the people suffering under the AI numbers kicked back and it started to look like the early stages of a full-on revolution and has become all they talk about in terms of data centres in the US.

And this mad, toxic mess is now being dropped on the UK and Scotland and the politicians don't seem to have learned or understood (or frankly even been aware of) all of this context. The Scottish Government can't hear any words that come before or after the words 'Inwards Investment', so they're in for a shock because this has the capability of being the big one.

I don't make social predictions, but there is a camel's back and there are a lot of straws on it already and something is going to be the last one that breaks it. If it happens and I had to bet, there would be worse bets than on it all beginning with a data centre. That's why this is such an enormous issue. That's why the political paralysis is especially worrying. And that's why you should get to grips with this.