Behind nearly every click and swipe you make is a network of energy-hungry computer servers and data centers. That’s because cloud based services, and especially the new world of artificial intelligence, require lots and lots of power.
Just last year, a Dallas-based company called Compass Datacenters signed a $3 billion contract to build rooms containing power management equipment. And they’re not alone.
According to recent reporting by the Wall Street Journal, data center developers are in the midst of a building boom. Bob Tita covers manufacturing for the Wall Street Journal, and he joined the Standard to discuss the trend. Listen to the interview above or read the transcript below.
This transcript has been edited lightly for clarity:
Texas Standard: Well, when we’re talking about these data centers, how big are they typically and how do they look and sound?
Bob Tita: Well, these are generally big, big places. They kind of look like big warehouses. And they’re very secure places.
So generally they’re kind of broken up into a series of big rooms where you would have, typically, hundreds of servers running. And as I said, they’re very secure and those rooms have, you know, there’s separate cooling and separate power sources and they’re typically for big customers – you know, big tech companies. That sort of thing.
Well, in your story, you talk about the cost of cooling the machines that process all that data. Can you say a little more about that?
You know, managing heat is a big issue for data centers because the servers, and the chips themselves, generate a lot of heat when they’re working. So it’s kind of like when we work out, we sweat. That’s kind of the way these servers are, too. And so managing the heat in these rooms is a very big part of being a data center operator.
And so it can become very costly to sort of cool them with a traditional A/C system like we think of in our homes or in a store or something like that. So the data center operators use different ways to cool these places down that are as efficient as possible.
So from your reporting, can you tell us what kind of companies and what kind of components are going into these data centers?
A lot of them, the components, are, of course, cooling. You want to cool down these rooms where the servers are, plus you have to backup all the power that is being consumed by the the servers.
So you have companies that make generators – Caterpillar, Cummins, those kinds of companies that make big generators – most of them diesel powered. And then electrical equipment makers – like Eaton, Schneider Electric, ABB – they provide things like batteries, backup batteries and transformers… power distribution equipment.
Have things like ChatGPT and other easily-available AI tools accelerated the demand for energy?
Yes, they have, because these sorts of programs generally collect massive amounts of data or images. And in the process of doing that – when you’re creating what they call “training models” – you’re consuming huge amounts of electricity to do that.
So as they sort of draw in all this information or all these images, that takes up a lot of energy.
Well, in your reporting, you cite a figure showing us data center space expanded by 26% last year. Where are these data centers popping up?
They’re largely in places that have a lot of energy capacity.
So it’s typically in sort of areas that are closer to kind of big metropolitan areas. So Washington D.C., sort of Northern Virginia, is sort of the hub of the data center industry, but also places like Dallas, Houston, even Chicago, where there’s a lot of electricity available to operate these things.
Well, you mentioned Dallas and Houston. Do you think Texas and Texas-based companies will be playing a big role in this boom?
Sure, because there’s a big tech industry in the state. And there’s a lot of power capacity.
Well, looking forward, do you see any sign that this trend will slow down or will continue to just accelerate in the near future?
It will probably accelerate for a number of years. It could slow down at some point, once you have the sort of build out complete.
So if you think about these AI sort of models, and once you have these huge data bases in place and created, using them doesn’t take maybe as much energy as building them. So, you know, in some years, at some point in the future, there’ll probably be kind of a leveling off a little bit.
But we’re not there yet. You know, this is going to go on for a number of years.
