Tech Overflow
We're Tech Overflow, the podcast that explains tech to curious minds. Hosted by Hannah Clayton-Langton and Hugh Williams.
Tech Overflow
Inside The Data Centre Boom
Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.
Data centres used to be the quiet “plumbing” of tech. Now they’re a frontline issue for AI, energy grids, local politics, and even election outcomes. We unpack what a data centre actually is (a giant, secure warehouse of compute, power infrastructure, and cooling), why hyperscale cloud providers like AWS, Microsoft Azure, and Google reshaped the way companies build software, and why the world suddenly feels the impact when a major outage hits.
From there we follow the AI trail. GPUs have become the workhorse for modern machine learning and large language models, but they also bring a brutal reality: more computation means more electricity and more heat. We dig into what “power is the bottleneck” really means, how grid connections can lag demand, and why some operators turn to on-site gas turbines and diesel backup despite sustainability targets. We also talk about the less glamorous engineering that makes AI possible, including water cooling, closed-loop systems, and why warm water out of a data centre can sometimes heat homes in places with district heating.
Then we zoom out to the human side. Journalist Dan Ziffer joins us with a vivid story from Australia, where local resistance to new transmission infrastructure and data centre expansion shows how quickly tightly priced projects can wobble. When communities carry the noise, water use, and loss of amenity while seeing few lasting jobs, regulation stops being abstract and becomes inevitable.
If you found this useful, subscribe for more clear tech explainers, share the show with someone who keeps hearing “the cloud”, and leave us a review with your take: should data centres face tougher rules where you live?
Like, Subscribe, and Follow the Tech Overflow Podcast by visiting this link: https://linktr.ee/Techoverflowpodcast
Dan Ziffer They are absolutely furious, and now they've worked out Dan Ziffer actually take that very many furious people to stop these Dan Ziffer because they're so highly geared, they've been built out so Dan Ziffer that even the tiniest grain of sand in the gears will stop the Dan Ziffer whole machine.
Hannah Clayton-Langton:world and welcome back to the Tech Overflow podcast. As always, I'm Hannah Clayton Langton.
Hugh Williams:And I'm still Hugh Williams, and we are the podcast, as know, that explains technology to curious minds. How are you doing, Hannah?
Hannah Clayton-Langton:I'm good. Um, I was gonna say we're the podcast that explains technology to curious minds, and we are the podcast that is always on move because I know that you are dining in from a different country. I am indeed I am.
Hugh Williams:I'm in the little uh little room out the back of my place Los Angeles. Hannah, and where are you?
Hannah Clayton-Langton:I am actually in London, so a little bit of stability there from me. So today we're talking about data centers, which is just topic of like debates, dinner party conversation. If you've followed the US midterms, which was really hard to miss, data centers were relevant there. And if you were anywhere near a news source on Friday, outage.
Hugh Williams:21 hours. 21 hours. Then there's Jensen Huang resigning as CEO of uh the world's most valuable company and huge chipmaker, AI chips, you know, the fabric that sits inside a data center. So uh data centers, data centers, and more data centers have.
Hannah Clayton-Langton:It's it's crazy two things. One, that that Jensen Huang news to me already feels like news because so much happened at the back end of last tech, and guys will cover that in the next this month in tech for sure. But also that those topics, that the common thread with US midterms, Microsoft outage, NVIDIA share prices, bananas and Chinese espionage, honey pot, like all of these things linked by data centers. So before we get into the inevitable community backlash, relevance, et cetera, like as we are tech overflow, let's start with the basics. to strip it back, Hugh, what is a data center?
Hugh Williams:Well, I think the best place to start is just a big broad that's the physical infrastructure that supports the world. So whenever you're using an app on your phone, connecting to something with your web browser, using a productivity tool work, most likely something is going on in a data in the world that's supporting what you do. So they're really the physical computing infrastructure underpins almost everything we we do today.
Hannah Clayton-Langton:Okay, so they're sort of like the the plumbing of tech.
Hugh Williams:Yeah, that's it. And a little unglamorous like plumbing, I think, Hannah and plumbing involved. We'll come back later on, perhaps, and talk about power and cooling. But uh, you know, they're big warehouses, basically data very large warehouses, very anonymous-looking warehouses. And they're, you know, they're really full of three basic some power infrastructure, because these are pieces of infrastructure, lots and lots of computing you know, computers, CPUs, GPUs, racked up in you know that look like big kitchen cabinets, and maybe later on talk about how they're built. And then cooling infrastructure, because these chips very hot, particularly the GPUs that underpin the that's used in AI. These run very, very hot, and so there's also cooling but great big warehouses, enormous warehouses, you know, full of computing equipment and the things to power and cool
Hannah Clayton-Langton:Okay, and if I have understood everything you taught me so this is a good test. Now that we live post the advent of cloud computing and so things are existing in the cloud, we have these centralized warehouses of like computing and data processing power, before they might have been like more local servers.
Hugh Williams:Exactly right, Hannah. Um, this podcast is going well. Um, you know, back when I when I first became a professional, the only computing infrastructure you had was the infrastructure that was sitting in front of you. So, you know, desktop computers, the advent of laptops. Um even the early phones really weren't connected to right? They weren't that smart. But as times moved on, that infrastructure's kind of moved the desktop to other places. And first of all, it moved into rooms in the office. So, you know, you used to be able to walk around your office and there was a server room. You went in there, lots of blinking lights, really cool, air conditioned very, very well. Some computers inside there, and so the sort of the that supported the office that you worked in was in a server room somewhere. And then, you know, over the last 15-20 years, those server rooms have largely disappeared. They don't have computing infrastructure in them anymore, And that infrastructure's moved off-premises, first of all, into warehouses that were often owned by large companies, and in more recent times into infrastructure that's owned by the big three cloud providers. So that's Google, that's Amazon's AWS, and that's Microsoft. And there's lots of sort of so-called neo clouds and all of other companies offering infrastructure, but it's those three big gorillas that provide most of the computing infrastructure that's in data centers that powers most of the things that we do in our digital lives.
Hannah Clayton-Langton:and before we inevitably segue into AI, because that is to every Tech Overflow topic these days, data centers a new concept, right? Like I did a little research before the show, and Google its first hyperscale data center in 2006. So that was 20 years ago, and I think AWS followed sort of thereafter. So, like the concept of them not new, but maybe the demand them ever increasing.
Hugh Williams:Yeah, that's it. You know, I was at Microsoft in 2006, and we were building data centers, filling them with tens of thousands, maybe hundreds of thousands of computers at that time, was really just at Microsoft for the use of the teams and the same was true at Google. You know, it was probably a few years later, probably in the early sort of 2010s, where those companies realized that infrastructure could be used by others and that what they'd built for themselves was actually a product that could be But certainly, yeah, you know, 2006-2007, you know, folks Amazon and Google were busy building their own data They coined this phrase hyperscale, and it really just meant that there was more than 5,000 computers in the data center. So that's kind of what that means.
Hannah Clayton-Langton:Okay, I'm interested, maybe we can get later to like how computers are in some of the big ones that we're talking But why data centers are so often in the news at the minute is this link with AI, right? Like you mentioned GPUs. Maybe we can remind listeners what that all means, but GPUs means thirst for processing power, right? And that is all powered by data centers.
Hugh Williams:That's right. So yeah, if you look back in history a little bit, you know, even just go back five years, the majority of the they often just call them servers that were in data centers were just full of CPUs. And CPUs are the general processing unit that um that you'd have in your laptop, that you have in your phone, it's the of thing that that runs a general purpose piece of computing. But in more recent times, with the advent of the large models and and the you know enormous scaling of those, a of the chips now that are in the data centers are GPUs, and these are originally graphics processing units. So they're particular chips that are designed to do many, small operations in in parallel, which means at the same So they're really like a really smart maths unit to do in parallel, and that turns out to be the kinds of chip you need to do the fundamental computation that's used AI. And those chips consume a lot more power and generate a lot more heat than the CPUs.
Hannah Clayton-Langton:Okay, so we need a lot of data. Is it data or is it processing power? It's sort of both.
Hugh Williams:It's sort of both. I mean, look, cloud computing is a lot about data, So when you're you know when you're using sort of at work or on your phone, it's a lot about storage as well. So, you know, Google Docs or uh, you know, Microsoft Word or your email or whatever it is consumes a lot of storage. And so, you know, traditional cloud computing is about power, typically CPUs and a lot of storage. But when it comes to AI, it becomes a lot more about and GPUs. So much more sort of chip-heavy, computation-heavy uh and workloads.
Hannah Clayton-Langton:I saw one stat, albeit from 2024, and the world moves so you can tell me how much you think it's aged. But back then we were generating 400 million terabytes of around the world every single day. Like, not for the whole year, every day.
Hugh Williams:Yeah, that's right. I mean, enormous amount of data. We need to just think about all of the social media, those of things, um, you know, training data for LLMs, all of these kinds of things is absolutely an exponential explosion amount of data that's stored and used. And again, that's all stored in, you know, the data that are largely built by these large companies.
Hannah Clayton-Langton:Okay, so 2024 stats are probably well out of date because I think the primary use of data centers is set to be AI, right? And growing.
Hugh Williams:Yeah, that's it. And that's what's really driving the power usage and the challenges. And I we'll come back to those and talk about those, Hannah. But that's really the part of data centers that's exploding right now, is the construction of data centers to support applications in AI.
Hannah Clayton-Langton:Yeah, and I guess their pull on those resources is again why they're relevant to communities and elections, et Okay. And you've been to a data center, I think, at least 10, 15 ago, is that right?
Hugh Williams:Yeah, that's it. It's hard to get permission to go to a data center, you're a tech exec. They're very, very secure facilities. Um, they typically only have a few tens of people working in them. So they're very software uh operated, you know, they're very sort of steady state. They just sit there and they run. And you know, large companies like to keep those very, secure because, of course, we've talked about the kind that's that's stored and the kind of workloads that are run in these data centers, and you don't want people just into those data centers and having access to that data. So they're very, very secure facilities. Um, so you have to ask very, very nicely, even as an to get permission to go into one, and then you you know you get escorted around the data center and you get to take a But um yeah, I've been lucky enough to be in a few. But that's not typical for somebody who works in uh West tech here in the US.
Hannah Clayton-Langton:And while we're in the sort of physical element of the data centers, it kind of throws me back to season two, I think it was, where we talked about AI data, because the for me, I guess, matters for two reasons. Like one, we talked a lot about um cooling and they generate a lot of heat. And I read that a lot of the tech companies are putting them in like cold locations, so like Sweden and maybe Denmark, I actually can't remember, definitely Sweden, and which makes sense because it costs less to cool something down when when it's cold out. But then at the same time, if I remember correctly, your data is being like physically processed is where the rules apply. So it doesn't matter where your company's based, if the is being processed in Sweden, for example, it's going to subject to Swedish regulation and law around that data. Is that right?
Hugh Williams:That's right. That's right. So, you know, these are physical pieces of infrastructure exist within countries, and so they have to follow whatever laws apply within those countries. And so you often find that countries are concerned about data is stored, and of course their governments are, um, so are large corporations. And so that definitely is one factor that goes into thinking about where to build data centers is where exactly is this going to occur and where exactly is this data going to be as they say, at rest.
Hannah Clayton-Langton:Okay, and just to take us back, like as someone who never even went into the server room in an office, because I'm not gonna lie, guys, I maybe wasn't super interested back at the time, but like this is like a big empty, is it just like a normal warehouse? Do you have to have special adaptive provision? Like, what does it actually look like? Is it like a ghost town of just like whirring and beeping flashing lights? Like, I don't know, just paint a picture.
Hugh Williams:Uh they're very soulless pieces of infrastructure, honestly, honey.
Hannah Clayton-Langton:not helping the PR.
Hugh Williams:No, it's not. No, it's not. No, it's a little bit uh it's a little bit Death Star kind of uh Star Wars type type of infrastructure, if I'm honest. So, you know, they're very, very large warehouses with you these almost endless corridors, fluorescent lighting, a of blinking lights, sort of the whir of fans, but you know, really no people, and really just very large pieces of warehouse with stacks and stacks of computers inside them. So, you know, really just designed to squeeze as many in as you can, get power to those computers, get cooling to those computers, and you know, run those in a big secure
Hannah Clayton-Langton:So they have to be presumably provisioned with like and plumbing than your average warehouse on the side of the road?
Hugh Williams:Yeah, absolutely. I mean, look, there's probably a few different things you about when you're building a data center. First of all, I guess is how close is it to the people that I want to serve, right? So if you're in tech, you'll often hear people talking the Amazon AWS data centers that are on the east coast US. They're very popular. And the reason they're very popular is they're very near US populace, and they're also not too far from Europe, right? And there's fiber optic cables that run under the Atlantic the US and Europe. And so, you know, you can get a pretty quick turnaround of from the east coast of the US. So if you're a large company, you want to put some data data center, you want to run a service from a data center, pretty interested in the east coast of the US. You're also interested in the west coast of the US, you might be a little bit interested in Europe, you're interested in Asia, depending on who you are. Um, very uninterested in Australia, by and large. Uh, you're so like, oh, we can serve that out of Asia, maybe West Coast US will be fine. And so it's a lot to do with sort of physically how close infrastructure is to the customers you want to serve, is one factor. Another factor is power. Um, that's a really, really big deal, is how how can I to these very hungry pieces of infrastructure. Third, you know, cooling. Fourth, probably, you know, how cheap is the land, what's tax jurisdiction, you know, those kinds of things. So lots of things to kind of think about.
Hannah Clayton-Langton:Okay, we need to put a pin in the fiber optic cables under Atlantic Ocean because I want to know more about that on a Um, but I guess like to state the obvious, I presume they generator backup and like a backup for the generator backup and stuff like that. Because like when the data centers go down, we notice. Like, I even I can't remember when it was, I think it was 2021, 22. Like there was a big outage at the company I was working at the time, and it was nothing to do with us. It was because a data center somewhere on the West Coast, East Coast it was of the US, was down and it was.
Hugh Williams:it's the AWS East.
Hannah Clayton-Langton:Exactly.
Hugh Williams:Yeah, that's right. mean, look, you know, the the way these things typically work is that there's some external power source, right? And we can come back to talk about different types of power source and things, but that's what largely runs the They then have inside the warehouse some form of battery that's going to allow it to run for some period of time, not very long, for some period of time. And then they also typically have some form of could be powered by gas, could be a gas turbine, could be a diesel generator, whatever it is. And so typically you've got three things. And typically what happens is when the power gets cut, external power gets cut, the batteries can kick in very, very quickly, you know, probably sub one-tenth of a second, nothing gets disrupted. But those batteries don't last long, right? Because you've got you've got a huge amount of infrastructure you're running, but it's enough time for you to be able to up the gas turbines or fire up the diesel generators so they can take over powering the data center. And you'll typically be able to get many hours of power out of those pieces of infrastructure. And the chances of the grid going down for such a long of time that you completely run out of diesel for generators or gas for your gas turbines is very, very low. So typically it'll go external power cuts over to the and then to some other power source that's powering the center.
Hannah Clayton-Langton:I remember from our chats with Maz, so I think that was six in season two, Ramaz Nam came on to talk to us all about the future of AI, but I'm sure he said that like the scale-up was one of the biggest bottlenecks for like AI Is that literally like grids being able to provision enough power to run all the different data centers that we need? When we know like physical, physical development is lot harder than um software pure development because it takes more time to like build things, get supplies, etc. So is that you think that's still true today?
Hugh Williams:Yeah, that's very, very true. I mean, power is the limitation in most cases. You know, it's not that hard relatively to build a Um you can get hold of the computing equipment, the network cabling, you know, the copper that you need to deliver the is, but it's actually getting power connected to it that deliver sufficient power to power a data center, that's the limiting factor. So, what sadly we're finding today is that a lot of these companies are building gas turbines at the data centers to the data center because they simply can't get enough power the grid. So the grid infrastructure is just not keeping demands of data centers. And because these companies need these data centers to the services they want to deliver, they're just building gas turbines, putting those on the data centers and running them using you know fossil fuels, which is pretty sad. Um, but it's really the you know the electricity grids holding them back.
Hannah Clayton-Langton:And is it true, Hugh, that like some companies are using at least leveraging a little bit of like nuclear power as
Hugh Williams:Yeah, that's right, Hannah. I mean, nuclear is definitely part of the powering I mean, I had a look at some statistics. It looks like roughly now 30% of the power to data centers is provided by coal power stations, 26% by gas, 27% by so that's the bit we can cheer about, and then about 15% by nuclear. So it's a little bit over half fossil fuels that's powering data centers today. And in the US, gas is the single biggest source at about 40%.
Hannah Clayton-Langton:Okay, okay. Well, some of that is a little bit icky. Um, and when we talk about ability to scale quickly, and think we should spend too long on this because I think it probably a topic in its own right, but like getting hold of the GPUs themselves, that can be a limiting factor too, right?
Hugh Williams:That's it. And all of these companies have enormous orders in and folks like NVIDIA for GPUs, and they often have to wait. It's a little bit like ordering a, you know, a jet for your airline, you know, join the queue. Uh, you'll get your jet when your jet's ready. But I still think the limiting factor today for most of companies is power.
Hannah Clayton-Langton:moving us away from power and focusing a little bit more on cooling, as we know, the chips themselves generate a huge of heat, and something's got to cool that heat down to stop the whole thing from just like cooking itself, right?
Hugh Williams:Yeah, that's it. And you know, if you if you sort of think about how you cool something, right, the most obvious thing to do is to well, let's open it up and have air blowing through it. Turns out that doesn't work very well and obviously isn't secure. The other obvious thing you could do is install air and that's traditionally how data centers were cool, but that turns out to be not super efficient either. It turns out the best way to cool data centers is actually water through them. And so modern data centers typically use water to down the chips that are actually inside the servers inside data centers.
Hannah Clayton-Langton:and that's like basic sort of physics question. Like water is a good like vector of temperature, right? Like it'll pick up the heat and then they pump the water of the building and the heat goes out with it.
Hugh Williams:Yeah, that's it. And you know, if you want to sort of think about things in terms, well, you know, water carries away roughly 3,500 more heat than the same volume of air. So it's a really, really efficient way to cool down a data And the way it works, actually, Hannah, is it's it's plumbing, right? So these data centers are typically built in modules off-site. So they're fabricated somewhere, and typically they're a couple of kitchen cabinets, if you like, the actual racks. So think of a think of a couple of kitchen cabinets facing other. They've got slots where you can slide in the computers. They might be put into the racks uh off-site, but then a little bit fragile when you move them, or maybe the are put in on-site where it's a little bit less likely you'll damage them in transit. But anyway, back to these kitchen cabinets. These kitchen cabinets have a big copper rail that runs the top of them, that's where you connect the power to each one of these kitchen cabinet racks. And they also have plumbing. So there's basically a cold water in and a warm water out goes into these racks, runs down the rack, and then runs each of the computers, and literally the water runs over top of the GPU or the CPU. So there's actual water circulating over the top of the that are deep inside the computers that are racked inside the server racks, and basically it's cold water in, warm
Hannah Clayton-Langton:it right that in some of these countries like Sweden, Finland, they're using the hot water out to heat homes?
Hugh Williams:Yeah, that's right. One of the unique things about Scandinavia, and I know lots of our listeners will have been there, is that there's infrastructure for heating houses. And so a lot of people live in large apartment buildings, efficient to heat, um, and the town actually pumps warm through the buildings, right? And they have these big, you know, heating facilities that heat up the water, pumps it around the town. And so you can hook up a data center to that. The only thing is the water's only warm when it comes out a data centre, it's not hot. Um, so you might have to heat it a little bit more, but you know, it's one of the great side effects of a data center is they produce warm water and you can actually use that heat towns. But I think that's very much a Scandinavian phenomena. Obviously, in countries like Australia and and many other around the world, we don't have that kind of
Hannah Clayton-Langton:that is classic scandy life. Okay. And the water itself, I think, has to be fair, isn't this of the challenge? Is it has to be fairly pure, like potable, like drinking type water? Is that right?
Hugh Williams:Historically, yes. And, you know, if you can get potable water, so drinking um, then it's cheaper to build the data center because you need you don't need things like stainless steel that are to corrode. But today there are data centres being built that use and water that has other impurities, but the cost of building the plumbing inside those data centres is much more than if you've got potable water available. So right now, seawater and impure water is sort of an edge but I can see a world Hannah over the next little while data centres are built so that they can, you know, use water of lots of different types.
Hannah Clayton-Langton:Interesting. Well, I hope they're not putting the warm seawater back in oceans, because that creates a different problem. But I'm sure there's plenty of smart mines trying to climate aspect of that.
Hugh Williams:Some of them are actually recirculating systems now. So in the old days, you'd you'd literally pump the water from the river into the data center and it'd circulate around, pump the warm water back out into the river. Today they're also working on ones that are a little like a car radiator, right? Where you fill the data center up with water and the water pumps around and it's re-cooled and then pumped around So they're a closed system that doesn't take water from or give water to anywhere. So lots of innovation going on in the cooling as well.
Hannah Clayton-Langton:Interesting, interesting. Well, I know that um Maz, who I mentioned earlier, and uh John Liddy, who was in episode 10, I think it was, of season two, both of them are super smart guys, and they're both really on the climate tech space right now. And I imagine that if you're in climate tech, this is a big, juicy, valuable problem to solve.
Hugh Williams:Yeah, absolutely. And it's certainly one that's in the public you know right now, too, Hannah. You know, I think lots of communities are rebelling against the fact that you know these data centers are being built their area and using lots and lots of their electricity and lots and lots of their water. And so I think there's a real incentive for investors to in companies that improve the electricity situation for data centers and improve how data centers are cooled. So definitely, definitely a hot area for investment.
Hannah Clayton-Langton:Okay, and on the topics of community backlash and politics. Actually, Hugh, I was super lucky. I was able to sit down earlier this week with Dan Ziffer, actually a journalist at the ABC, which is Australia's public broadcaster, like BBC here in the UK. And Dan covers all sorts of uh business issues. So, of course, data centers are a big topic for him And he's gonna kick us off with a super interesting um, a company called NextDC, which is an Australia data company, which sort of highlights some of the biggest of this debate. Dan Ziffer Well, there's been some really interesting things happening Dan Ziffer Australia, which uh, you know, I'm aware is uh for a lot of Dan Ziffer or the other side of the world a really small population. Dan Ziffer Uh, but there have been some real flashing red lights about Dan Ziffer and the data center build-out that have come from here. Dan Ziffer The first of these was Next DC, which is a veteran data center Dan Ziffer operator in Australia. Dan Ziffer They've got about 19 sites. Dan Ziffer They're building a lot more. Dan Ziffer Now, they had one site in the inner west of Melbourne, where I Dan Ziffer live, and it was bubbling along. Dan Ziffer You know, the community wasn't very happy about it, but you Dan Ziffer they're often not about data center uh development. Dan Ziffer What happened was a street a couple of hundred meters away from Dan Ziffer an existing site where Next DC wanted to build a much larger Dan Ziffer next to it, saw the march of industrial-scale power poles Dan Ziffer them. Dan Ziffer Now, in Australia, suburban power poles are generally made of Dan Ziffer wood and they're about seven meters high. Dan Ziffer These were more like 15 to 22 metres high. Dan Ziffer They're made of concrete and they can obviously deal with Dan Ziffer larger amounts of electricity. Dan Ziffer Now, this street saw the march of these poles, saw the Dan Ziffer holes that have been dug in their street uh to accommodate them, Dan Ziffer and said uh in a very Australian way, nah. Dan Ziffer Um they parked their cars in the street, they moved their bins Dan Ziffer into the street. Dan Ziffer The electricity distributor that was doing the work uh saw this Dan Ziffer kind of civil disobedience protest, reversed out at full speed, Dan Ziffer and then uh announced it would begin a kind of community Dan Ziffer period. Dan Ziffer So that was bubbling along. Dan Ziffer About three weeks later, NextDC, which is a listed company Dan Ziffer about $7 billion, put out a notice to the stock exchanges Dan Ziffer required to do in Australia, basically saying that Dan Ziffer uh would delay the construction of their massive site. Dan Ziffer It was gonna cut one-third of their projected future earnings, Dan Ziffer it was gonna push them back by about a year. Dan Ziffer They were gonna start selling land that they had banked Dan Ziffer developments, they were gonna stop doing any developments Dan Ziffer that didn't have a set customer, um, absolutely smashed the Dan Ziffer the share price of which fell 42% in a day. Dan Ziffer And really, what it showed is in a lot of the valuations of Dan Ziffer AI or AI adjacent companies, everything is priced to Dan Ziffer Everything must run as the company has suggested for the Dan Ziffer to be justified. Dan Ziffer And in the real world, that's just not how things work. So the point here, Dan, that's really interesting. I guess it's twofold. One is the sensitivity of some of the valuations that sit these companies, which, as you rightly explained, is naive, but also the impact that community groups can stalling development with things like local protests, right? Dan Ziffer Well, what they've learned from this one is you don't even need Dan Ziffer to stall it very much or for very long. Dan Ziffer Everything is so highly geared and so tightly measured that Dan Ziffer a small delay could mean the project doesn't happen because Dan Ziffer numbers will no longer stack up. Dan Ziffer Now we've seen this in recent years in our regional areas. Dan Ziffer Australia is going through an immense energy transition. Dan Ziffer We are getting off coal, largely because the plants that burn Dan Ziffer it for power are extremely old. Dan Ziffer They're beyond their lifespan. Dan Ziffer And so it's uh not in some ways an ideological debate. Dan Ziffer It comes down to physics and economics. Dan Ziffer We are well down the road of an energy transition. Dan Ziffer It is by no means certain we might not make it. Dan Ziffer But one of the big elements of it has been the construction of Dan Ziffer high-tensile transmission lines through regional, rural, Dan Ziffer land. Dan Ziffer That's been very controversial. Dan Ziffer And the data centres are quite similar because what you do is Dan Ziffer you have communities that are going to bear the cost of this Dan Ziffer infrastructure, being a loss of amenity, um, increased Dan Ziffer risk, uh a loss of land value, those kind of things, but Dan Ziffer not going to enjoy the benefits uh that it brings, such as Dan Ziffer stability, um, you know, a more thriving economy, that kind of Dan Ziffer stuff. Dan Ziffer So this has been bubbling away in regional areas. Dan Ziffer It has been a big issue. Dan Ziffer Now it's essentially come to the city where you have denser Dan Ziffer and you have sites that are extremely large, that uh you know, Dan Ziffer give off a never-ending hum of sound, uh, that have diesel Dan Ziffer that uh get checked a lot and might have an Olympic-sized Dan Ziffer pool of diesel fuel just sitting there in case it's required. Dan Ziffer These are not great neighbours for residential areas. Dan Ziffer And the other thing about them, when you go to them, and many Dan Ziffer of the listeners will have, they're kind of dystopian in that Dan Ziffer they're devoid of workers. Dan Ziffer Now, there's a huge amount of uh construction and uh lots of Dan Ziffer in the construction process. Dan Ziffer But once data centers are finished, they employ very few Dan Ziffer in and of themselves. Dan Ziffer You know, someone had turned on the robot vacuum planner. Dan Ziffer So they have become this flashpoint because uh even as the Dan Ziffer of Next DC, uh Craig Scroggy says, Um, they are the Dan Ziffer of people's fears about AI, about a really uncertain future. And these debates are having real impact on, for example, US midterm elections were last week at the time of recording, right? They're having a real material impact on where things are Dan Ziffer Oh, they absolutely are. Dan Ziffer Now, for a long time, data centers were small in our industrial Dan Ziffer areas and they were for storing information. Dan Ziffer You could drive past, you wouldn't know what they were. Dan Ziffer It looked like a warehouse. Dan Ziffer As they got bigger, and as the requirement for them to be Dan Ziffer to customers meant that they got larger, they consumed more Dan Ziffer they consumed more water, they became more visibly prominent, Dan Ziffer um, people became more concerned about them. Dan Ziffer Now, any local community where you're gonna have a loss of Dan Ziffer uh, you know, you're gonna get people really worked up about Dan Ziffer As these sites moved out of industrial areas and into more Dan Ziffer residential areas, there was always gonna be that inevitable Dan Ziffer But this isn't a PR problem. Dan Ziffer It is in one sense, and they haven't been able to sell the Dan Ziffer but it's a bit like the transmission lines we discussed, Dan Ziffer the people who are going to feel the loss of amenity, who are Dan Ziffer going to have light pollution, noise, a big long Dan Ziffer project, a permanent large black box outside their house, Dan Ziffer real, right? Dan Ziffer That is a real loss of amenity. Dan Ziffer It's like you know, your neighbor putting up a massive Dan Ziffer wind turbine. Dan Ziffer You will notice it, right? Dan Ziffer You have a loss of amenity. Dan Ziffer Your property value will go down. Dan Ziffer You probably won't feel the benefits. Dan Ziffer What the industry probably hasn't done very well is explain the Dan Ziffer benefits, but also maybe talk to some of their uh proponents Dan Ziffer say, is this really the spot? Dan Ziffer Is this the spot? Dan Ziffer You know, we have the driest continent on earth, and there Dan Ziffer talk about us being an AI data supercenter. Dan Ziffer Now, maybe in certain areas where we have abundant renewable Dan Ziffer and we have lots of water, that could be a really good thing. Dan Ziffer But in areas where power bills have gone up immensely because Dan Ziffer we're essentially trying to squeeze through this narrow pipe of Dan Ziffer an energy transition, which is very expensive to do but Dan Ziffer Of course, communities are going to be absolutely white hot Dan Ziffer what they see as large companies getting a free ride on Dan Ziffer amenity so that they can push their power bills up, take Dan Ziffer water, and leave them with this massive black box outside their Dan Ziffer house. Dan Ziffer That doesn't seem like a very good deal for them. Dan Ziffer And understandably, politicians have been unable to say, Dan Ziffer no, this is a good deal, there's going to be lots of jobs. Dan Ziffer These have become the physical manifestation of people's Dan Ziffer about AI, their uncertainty about the job market, and the Dan Ziffer in most developed nations that since COVID, people have been Dan Ziffer backwards. Dan Ziffer Their real wages, their wages uh minus the impact of Dan Ziffer are lower. Dan Ziffer So they are going backwards. Dan Ziffer Things are harder. Dan Ziffer And so when they see a sector that seems to ride rifshod Dan Ziffer concerns of people that seems to have a kind of politically Dan Ziffer run, they are absolutely furious. Dan Ziffer And now they've worked out it doesn't actually take that very Dan Ziffer many furious people to stop these things because they're so Dan Ziffer geared, they've been built out so fast that even the tiniest Dan Ziffer of sand in the gears will stop the whole machine. The political outcomes had real impacts on things like the valuations, right? Just straight off the bat after the blue wave we saw in the US. Dan Ziffer Yeah, absolutely. Dan Ziffer It's kind of like a welcome to the real world moment. Dan Ziffer I mean, you don't have large multinational food companies Dan Ziffer decry food labelling laws or food quality laws in the Dan Ziffer in which they operate. Dan Ziffer You know, you don't have your really big global firms Dan Ziffer that they can't use poisonous but cheap ingredients in Dan Ziffer that they sell. Dan Ziffer It's just something that they deal with as part of operating Dan Ziffer a sensible business. Dan Ziffer Uh, what we're going to see with data centers and maybe even Dan Ziffer the AI companies as well, is a return to reality in that you Dan Ziffer create a product that might have substantial harms without doing Dan Ziffer your best to reduce those harms. Dan Ziffer We have this in finance, we have this in car manufacturing, we Dan Ziffer have this in all these things. Dan Ziffer Red tape, uh, you know, that's a that's an argument you have Dan Ziffer but essentially when you talk about regulations, it's ways to Dan Ziffer keep people and things safe. Dan Ziffer And we're going to start to see the imposition of ways to make Dan Ziffer this technology safe for people and for the community, Dan Ziffer politicians have worked out that uh their people who vote for Dan Ziffer them every uh you know couple of years are really, really Dan Ziffer up about the unchecked nature of AI and data-centered Dan Ziffer They don't see the benefits, even if they experience Dan Ziffer their phones and listening to this uh podcast uh, you know, Dan Ziffer they're walking, but they don't feel the benefits and they see Dan Ziffer the downsides. Dan Ziffer They're really concerned.
Hugh Williams:That was super interesting, Hannah. And thank you, Dan, for joining us on the podcast. It was just great to have your insights. Uh Hannah, we had the US midterms just over a week ago, and I think all our listeners will know it was a so-called blue tsunami where the Democrats took control of Congress. And one of the things we're definitely hearing from the over the last week is that they want a national data moratorium starting now. And so something that swept them to power is actually this against data centers being built in communities, the of data centers and cooling. And so I think this is going to wash through the world over the next little while. And I would imagine there's lots of other countries this is great news for them as they invest in data center So there'll be more to this story, Hannah. Maybe something we can pick up in this month in tech in a episode.
Hannah Clayton-Langton:Yeah, 100%. It's just so relevant, which is uh kind of wild, but I it's that AI link, right, that makes data centers like somehow relevant to all of us.
Hugh Williams:Absolutely. I think that's a great way to bring the episode to a close, Hannah. Tech Overflow Podcast. You can find us on LinkedIn, X, and Instagram. And we also have our own website, techoverflowpodcast.com.
Hannah Clayton-Langton:That's right. And guys, if you've enjoyed the show and you like what doing, the best thing you can do to support us is to share the Tech Overflow podcast with your friends, with your family, your colleagues, um, who, to be honest, you know, they're interested in this stuff. Um, next week we've got a super interesting interview lined up all about cybersecurity, which is another endlessly in the news. So stay tuned, and we will see you next week.
Hugh Williams:you soon, Hannah. Bye.
Hannah Clayton-Langton:Bye.