The AI Buildout is a Boon for the Fossil Fuel Industry w/ Ketan Joshi

WEBVTT
00:00:00.000 --> 00:00:21.039
So the fossil fuel industry getting super excited, the tech industry carefully but clearly getting involved with the fossil fuel industry, and then the government trying to like get the public response of like thank you for you know regulating these scary, you know, like this scary new industry, but not actually regulating them properly.
00:00:21.120 --> 00:00:28.879
Um that is a really nice set of dynamics that I think other parts of the world can look at and be like, we do not want to do what Australia is doing.
00:00:48.640 --> 00:00:51.600
I'm your host, Paris Marks, and this week my guest is Katan Joshi.
00:00:51.759 --> 00:00:56.799
Katan is a climate writer and data analyst based in Oslo, working with climate and environmental groups.
00:00:56.960 --> 00:01:03.840
He's been releasing a bunch of fantastic reports recently looking at AI and data centers and so many different aspects of that.
00:01:04.000 --> 00:01:08.719
Of course, you'll find links to those in the show notes, but you know, we'll be discussing them in this episode as well.
00:01:08.959 --> 00:01:16.319
Katan was on the show, I believe it was earlier this year, where we were talking about the climate impacts and the environmental impacts of data centers.
00:01:16.480 --> 00:01:34.319
So I figured we would start by digging into that a little bit more because as you're probably aware, if you were listening to our data vampire series last month, my book, Hyperscale, comes out this month in October, um, which, you know, I would very much appreciate if you go pre-order a copy, comes out on October 20th.
00:01:34.480 --> 00:01:41.760
And of course, pre-orders really help authors and new books because it is a signal to booksellers that there's interest in this, right?
00:01:42.000 --> 00:01:46.959
It tells them that they should bring in more of these books because people actually want to buy it and read it.
00:01:47.040 --> 00:01:48.959
You know, it's a topic that people are interested in.
00:01:49.120 --> 00:01:56.640
So if you do enjoy this conversation, if you appreciate the work that I do, pre-ordering a copy of my book would really help me out and I would really appreciate it.
00:01:56.719 --> 00:01:59.359
And you can find out more by going to hyperscalebook.com.
00:01:59.519 --> 00:02:06.560
Now, beyond the climate impacts of data centers, we also talk about NVIDIA itself, you know, the chips that go into all these data centers, the climate impacts of that.
00:02:06.719 --> 00:02:12.879
And I also wanted to talk about a specific example of where we're seeing this data center build out happen around the world.
00:02:13.039 --> 00:02:17.199
And Katan, even though he lives in Norway, is originally from Australia.
00:02:17.360 --> 00:02:25.759
And so he has also been looking into the impacts of data centers in the country that he's from and wrote a fantastic report looking into that.
00:02:25.919 --> 00:02:39.439
And so I figured, you know, after we talk about the climate impacts, after we talk about the chips, after we talk about the impacts of data centers more broadly, that it would be cool to kind of zoom in on a specific location to see how this is all playing out.
00:02:39.520 --> 00:02:42.639
Um, and Australia was, you know, the perfect place to do that.
00:02:42.800 --> 00:02:47.520
It's also, of course, a location that we don't often talk about the impacts of data centers in.
00:02:47.680 --> 00:02:56.240
But as we're seeing in so many other parts of the world, the opposition to data centers and the proposals for data center projects are really growing there, as well as in other places.
00:02:56.400 --> 00:02:59.599
Many of them, of course, being powered by fossil fuels.
00:02:59.759 --> 00:03:02.159
So I think that you're really going to like this conversation.
00:03:02.319 --> 00:03:12.800
Over the next couple of months, I'll be kind of interspersing conversations about data centers with some other major issues, especially around AI and this discourse that we've been seeing around that lately.
00:03:12.960 --> 00:03:14.319
I do want to cover that on the show.
00:03:14.400 --> 00:03:18.000
So stay tuned in future weeks for those conversations to come out as well.
00:03:18.240 --> 00:03:20.879
All as we head into, you know, the end of the year.
00:03:21.039 --> 00:03:33.599
It's crazy to think it's September already, uh, where you know, we'll be looking back on the year and having some bigger conversations about the tech industry and, you know, the impact of tech on our lives as we uh close out 2026.
00:03:33.680 --> 00:03:36.960
Uh and, you know, who knows what 2027 we'll have in store.
00:03:37.120 --> 00:03:42.159
But as always, if you enjoy this conversation, make sure to leave a five-star view on your podcast platform of choice.
00:03:42.319 --> 00:03:45.759
You can share the show on social media or with any friends or colleagues who you think would learn from it.
00:03:45.840 --> 00:03:54.879
And if you do want to support the work that goes into making Tech Won't Save Us every single week and, you know, the work that I do more generally, you can go to patreon.com slash TechWon't Save Us where you can become a supporter.
00:03:55.039 --> 00:04:02.800
And of course, I would very much appreciate if you went to pre-order my book, Hyperscale, by going to hyperscalebook.com where you can find pre-order links and more information.
00:04:02.960 --> 00:04:05.439
Thanks so much and enjoy this week's conversation.
00:04:06.319 --> 00:04:08.719
Katan, welcome back to Tech Won't Save Us.
00:04:08.960 --> 00:04:09.439
Thank you.
00:04:09.520 --> 00:04:10.879
It's lovely to be back.
00:04:11.199 --> 00:04:11.599
Awesome.
00:04:11.680 --> 00:04:13.199
Uh it's always great to chat with you.
00:04:13.360 --> 00:04:23.600
You know, every time I talk to you, I feel like I learn so much because your insights on this topic and like the environmental dimensions of the tech industry are just like fantastic, right?
00:04:23.680 --> 00:04:27.279
Um, and so essential, which is why I keep wanting to talk to you and have you back on the show.
00:04:27.439 --> 00:04:34.560
And I'm sure why all these organizations keep wanting you to write fantastic and brilliant reports for them, uh, covering so many dimensions of this.
00:04:34.720 --> 00:04:48.800
You know, in this in this conversation alone, we'll be going over recent reports that you've written this year about the climate dimension of AI, what is happening with NVIDIA and and you know, kind of its climate impacts, and of course what we're seeing in Australia, right?
00:04:48.959 --> 00:04:55.519
A place that I feel like a lot of us aren't paying so much attention to, but where we see the data center fight really, really growing recently.
00:04:55.680 --> 00:04:58.639
And so I want to, you know, I want to get to Australia a little bit later.
00:04:58.720 --> 00:05:04.639
But earlier this year, you were on the show and we talked about the climate dimension of AI and of data centers, right?
00:05:04.720 --> 00:05:08.000
You know, the the kind of growing emissions, uh, what was driving that?
00:05:08.160 --> 00:05:14.639
And so I wanted to kind of pick up on that and to see how things have shifted over the past year.
00:05:14.800 --> 00:05:31.519
And so I wonder for you, when you think about these massive hyperscale data centers and the generative AI tools that they power, what stands out to you in terms of how we've seen that develop in 2026 and the real things that are kind of like red flags for you that we should be paying attention to?
00:05:32.319 --> 00:05:40.319
Yeah, I think definitely the obvious thing is just the size and the number of proposed gas-fired power stations.
00:05:40.560 --> 00:05:42.560
It's a quite a nice simple story, right?
00:05:42.639 --> 00:05:46.399
Like there's no weird grid power dynamics that you need to understand.
00:05:46.639 --> 00:05:59.360
They are literally just building a very large fossil fuel power station on the site of the data center and plugging it directly into it, and then just burning through a phenomenal amount of fossil fuels from the data center.
00:05:59.600 --> 00:06:02.399
And the plan is to build bigger and bigger ones, right?
00:06:02.480 --> 00:06:11.600
So compared to a year ago, there's actually a great um tracking project called uh Global Energy Monitor, and they track like oil and gas and fossil fuel projects around the world.
00:06:11.839 --> 00:06:13.759
And they make all this data freely available.
00:06:13.839 --> 00:06:21.600
Uh, and so what they do is they compare each half-year, they compare that report to the to the previous one just to sort of show what's changed.
00:06:21.759 --> 00:06:28.079
And easily the biggest change in the world in the you know, across like all of the planned power plants everywhere in the world.
00:06:28.480 --> 00:06:36.720
The United States and their plans for gas fired power stations increased by such a significant amount that it's just hard to believe.
00:06:37.040 --> 00:06:41.199
And many of those projects are explicitly for data centers.
00:06:41.519 --> 00:06:46.079
So some of them are on power grids that have a lot of data center development on them.
00:06:46.160 --> 00:06:50.480
So you can kind of attribute, you can generally pretty confidently say, well, that's because of the data center.
00:06:50.720 --> 00:06:55.759
But there's also a huge number that like they call, you know, whatever data center gas plant.
00:06:56.319 --> 00:06:57.920
You can you can sort of see the full list.
00:06:58.160 --> 00:07:01.439
And there's one that, you know, we'll go into a bit later when we talk about NVIDIA.
00:07:02.240 --> 00:07:04.240
But it's the second biggest on the list.
00:07:04.399 --> 00:07:13.120
Uh it's twice the size of the um current biggest operating gas plant in the world, which is a Saudi Arabian um gas processing facility.
00:07:13.360 --> 00:07:17.199
Uh and so like these mega projects are just getting worse and worse.
00:07:17.439 --> 00:07:22.480
Um and then the other thing that's worth noting is that I always like to track what's actually happening.
00:07:22.639 --> 00:07:27.600
So that that like one side is like the projections and the plans and the huge projects and that big list.
00:07:27.920 --> 00:07:30.639
But I'm always curious to know, you know, what is actually happening on the ground.
00:07:30.800 --> 00:07:33.759
Are they building them, you know, are they burning fossil fuels?
00:07:34.000 --> 00:07:38.639
I have to say I'm pretty surprised that many of these big projects have now entered into construction.
00:07:39.040 --> 00:07:45.040
So at the very least, we can say things are progressing as they hope, which is bad for us.
00:07:45.279 --> 00:07:47.519
You know, people who live in the atmosphere.
00:07:47.680 --> 00:07:59.680
Um that's not very good for us, but it's good for them because they are actually seeing the construction process commence for the you know, gigawatt scale data centers and the and the very large gas facilities that go with them.
00:07:59.920 --> 00:08:01.439
So uh that's really bad.
00:08:01.519 --> 00:08:05.600
And then and then, you know, there's kind of the grid side of things, which is a bit more mixed.
00:08:05.920 --> 00:08:11.519
There are some companies that have, you know, made big renewable energy investments, but they're a minority.
00:08:11.839 --> 00:08:24.240
There was a new report out just yesterday from an analytics firm called Energy Tag, where they just compiled a whole bunch of data center projects and they said, What energy source is this particular data center project linking to itself?
00:08:24.560 --> 00:08:33.279
And it and by my calculations, in terms of like energy generated, um the gas projects were six times the number of renewable projects.
00:08:33.600 --> 00:08:35.600
And so that's a terrible ratio, right?
00:08:35.679 --> 00:08:39.200
And like and the one I was just reading half an hour ago of Amazon.
00:08:39.519 --> 00:08:47.200
Amazon also has, you know, if you look at it in terms of just the capacity of the power stations, it looks like it's half and half, like half renewables, half gas.
00:08:47.519 --> 00:08:51.440
But renewables generate less as a proportion of their capacity than gas does.
00:08:51.759 --> 00:08:55.600
So it's actually more like a quarter renewables and three quarters gas.
00:08:56.000 --> 00:08:57.679
Roughly, it's probably a little bit worse than that.
00:08:57.759 --> 00:08:58.879
I mean they're generous.
00:08:59.200 --> 00:09:06.879
Uh and Amazon is like kind of weak, you know, today, literally today, they're saying, you know, they got asked, I think, for the first time, about their gas line.
00:09:07.200 --> 00:09:10.639
Uh this is like their first ever public statement about the gas line.
00:09:11.120 --> 00:09:20.720
And they basically just said, Oh yeah, well look, you know, we're not dropping our renewable energy targets, but we don't think that it's like a linear pathway to get to net zero by twenty fifty.
00:09:21.120 --> 00:09:24.399
And it's just like what do you think?
00:09:24.480 --> 00:09:26.080
Like, what do you think the pathway is?
00:09:26.159 --> 00:09:29.519
Because they're just going directly upwards on the chart of emissions.
00:09:29.759 --> 00:09:34.080
And so that's not I mean that's like exponential in the opposite direction.
00:09:34.320 --> 00:09:35.120
It's uh it's okay.
00:09:35.279 --> 00:09:35.759
It's okay.
00:09:35.840 --> 00:09:44.399
You know, what once the once all the direct air capture and and the carbon capture and sequestration comes online, that that line will come all the way back down, no problem, right?
00:09:45.200 --> 00:09:56.480
So so things have gotten worse, and not just in the sense that the pipeline's getting worse, but in the sense that we can see through satellite imagery bits of construction stuff being laid on the ground.
00:09:56.720 --> 00:10:05.519
Uh and I think a lot of people were skeptical, like, oh no, no, these projects are never really going to get built, but at the very least we can say they're on the way to being built, and that's not good.
00:10:05.840 --> 00:10:08.799
Yeah, and that was even something I was gonna ask you about, right?
00:10:08.879 --> 00:10:21.759
Because I feel like we have been seeing these stories for the past number of months that it's like, okay, they do have the plans to build out this capacity of data centers or of gas of what have you, but for whatever reason, they're being stalled.
00:10:21.840 --> 00:10:25.200
We're not seeing these things like enter into construction and and whatnot.
00:10:25.360 --> 00:10:39.440
But based on what you're seeing, you know, there's still a really worrying amount of construction that is that is ongoing that is picking up, which obviously, you know, theoretically would lead to the completion of projects that are going to be spewing out loads of new emissions, right?
00:10:39.679 --> 00:10:51.759
Yeah, and then the other thing as well to note on that is that a lot of the particularly the really large like multi-gigawatt gas plants, and so just to give a sense of scale, like a gigawatt gas plant is like to run a city or a state, right?
00:10:51.840 --> 00:10:54.799
Like we're not talking about like your neighborhood's you know, power station.
00:10:55.440 --> 00:10:58.000
These gas plants are being built in a modular way.
00:10:58.320 --> 00:11:06.639
So the idea is you kind of get like a quarter of it in the ground, and then you just see, you just take a look at your finances and you take a look at the situation, you're like, okay, now we're gonna build an exporter.
00:11:06.879 --> 00:11:13.120
So in a sense, that's that's like it's not as scary because you don't have one huge gas plant being built in one go.
00:11:13.600 --> 00:11:26.320
But in another sense, you could end up with worse emissions because like let's just say they kind of get halfway through and the bubble pops, you still end up with a really, really huge new gas plant that will need someone to use and to burn fossil fuels, right?
00:11:26.399 --> 00:11:29.440
Like, even if that's not a data center, it's some other uh thing.
00:11:29.519 --> 00:11:33.440
And this is actually a really great interview with Bloomberg, uh, Bloomberg analyst recently.
00:11:33.679 --> 00:11:38.159
I don't think they know that like we like activists like listen to this sort of stuff.
00:11:38.480 --> 00:11:40.399
Because they just say a lot of stuff out loud.
00:11:40.480 --> 00:11:44.320
I highly recommend listening to like, you know, gas analysts and stuff like that.
00:11:44.480 --> 00:11:46.879
Um, because they just really admit all this stuff.
00:11:47.120 --> 00:11:53.200
And and you know, the and this analyst was just like, yeah, yeah, we're not worried at all about like the data center bubble popping.
00:11:53.440 --> 00:12:04.720
Because we'll just be gifted with like this bounty of fossil fuel infrastructure, which we can then just use for like a power grid or some other industry, or like, you know, Bitcoin mining or something else, right?
00:12:04.799 --> 00:12:10.480
Like they're just like, we don't care, we've got all these gas pipelines and coal plants that have had their lives extended to five years.
00:12:10.799 --> 00:12:12.159
So they're not bothered at all.
00:12:12.320 --> 00:12:18.639
Um the data center companies might be bothered by the idea of the bubble popping, but the fossil fuel companies are like, we don't give a shit.
00:12:18.720 --> 00:12:24.480
We're just getting we're just getting a huge amount of money poured into fossil fuel infrastructure and we're gonna be totally fine.
00:12:24.559 --> 00:12:26.799
And they say as much, you know, when they talk about it.
00:12:27.200 --> 00:12:31.200
It's really worrying to hear, you know, exactly what you're saying there, right?
00:12:31.279 --> 00:12:40.480
That that they're like, okay, even if we don't have the data centers, that that's fine because we'll use all this new fossil fuel infrastructure for for whatever else, right?
00:12:40.559 --> 00:12:50.559
And that suggests to us that even if the data center AI bubble kind of implodes, we basically still have all of these new emissions effectively locked in, right?
00:12:50.720 --> 00:12:55.279
Because now this is on the grid or this is built and they're going to find a use for it.
00:12:55.360 --> 00:12:59.919
They're not just going to abandon all of this massive investment that they've made over the past few years.
00:13:00.159 --> 00:13:10.399
You know, as far as you are seeing right now, what is the climate and emissions impact of all of this gas that we're seeing built out in order to power data centers?
00:13:10.559 --> 00:13:17.039
And what does that mean for our hopes of hitting climate targets in the coming years?
00:13:17.120 --> 00:13:20.720
You know, so many countries had targets around 2030 and 2035.
00:13:20.960 --> 00:13:22.320
What does it mean for that?
00:13:23.200 --> 00:13:32.960
The United States uh greenhouse gas emissions, there's been a few analyses uh basically kind of updating the situation, right, now that all of these data centers kind of just emerge.
00:13:33.200 --> 00:13:38.720
Uh and I think each of these analyses takes, I would argue, a slightly too optimistic view.
00:13:38.879 --> 00:13:46.240
Like that they sort of downplay uh data center grid level emissions and gas emissions a little bit.
00:13:46.480 --> 00:13:54.080
But even then, uh, you know, you sort of get to see like they do like a normal scenario and then like a high data center scenario.
00:13:54.240 --> 00:14:05.759
And like the US, for instance, there's this sort of a between the four to five percent increase in the total in the country's greenhouse gas emissions, and the United States has the second highest grid emissions on the planet, right?
00:14:05.840 --> 00:14:07.360
Like just before China.
00:14:07.679 --> 00:14:14.080
It sounds like a small percentage, but that is like you know, country-sized uh amounts of greenhouse gas emissions.
00:14:14.240 --> 00:14:21.279
Um, and I would argue an underestimate, um, you know, just based on the dynamics that I know of how it happens in the US.
00:14:21.600 --> 00:14:36.559
One estimate, another one that I saw just recently was uh Bloomberg, you know, calculating the greenhouse gas emissions of all the planned data center-specific gas plants, of which I think there was like 95 or something, if I remember correctly.
00:14:36.799 --> 00:14:52.480
And it was like 325 megatons, which is um I have a I have a chart in the um recent report I released, comparing that to the total economy-wide emissions of several countries, that is about the same as all of the United Kingdom.
00:14:52.559 --> 00:15:04.159
So not just the power grid in the UK, but everything in the UK, all transport, all industry, all the power sector, all homes, that is gas-fired power stations to run new data centers.
00:15:04.240 --> 00:15:09.120
So not like you know, the whole data center industry, but just the ones that are planned.
00:15:09.279 --> 00:15:22.320
Um, and it's a testament to both the sheer absurd inefficiency of the software systems they're running, but also the weird desire, intense desire that they have to run it off fossil fuels.
00:15:22.559 --> 00:15:29.360
Those two things together, you end up with like numbers that are just so big that people find them a little bit hard to process.
00:15:29.600 --> 00:15:34.000
Uh because you're like, so that's adding in a United Kingdom worth of emissions.
00:15:34.080 --> 00:15:35.519
That's uh seems bad.
00:15:35.840 --> 00:15:36.320
Absolutely.
00:15:36.399 --> 00:15:44.879
It's it's like when you talk about, say, a cluster of data centers being like requiring a gigawatt of electricity, it can just feel so divorced from anything.
00:15:45.120 --> 00:15:53.919
But when you do talk about like all of these data centers needing as much power as the United Kingdom, it does just it does just sound completely absurd, right?
00:15:54.080 --> 00:16:07.679
Like, how is this even possible that this industry in the past few years is adding so or you know, planning to add so many more gas plants and so much additional emissions that that is really what we're facing, you know?
00:16:07.759 --> 00:16:19.519
And and I wanted to pick up on one other thing with you, which I think will help us to transition into, you know, looking at the broader supply chain and what you've been doing on NVIDIA, but also the claims that they make around AI, right?
00:16:19.679 --> 00:16:35.600
You know, that it's not just that, okay, you know, we're building out all this gas to power all of our data centers and this is just the way it is, but this kind of framing, this kind of greenwashing of what they're doing to say, yeah, we're doing this, but actually we're helping the climate.
00:16:35.679 --> 00:16:37.440
AI is going to be good for the environment.
00:16:37.600 --> 00:16:45.120
Can you talk to us about what you're seeing there and how they're actually trying to like make this argument work in any kind of coherent way?
00:16:45.679 --> 00:17:08.720
Yeah, we we uh I was I was um supported by a consortium of climate groups in February of this year, uh, and we had a real go at this idea uh because it was one that we saw all the time, particularly in like you know, not just sustainability reports, but uh companies were saying it to governments specifically to argue against themselves being regulated, and it was working.
00:17:08.960 --> 00:17:21.519
And and there's a story uh that Google would tell all the time, most recently to an EU policy uh submission process that AI can reduce the world's greenhouse gas emissions by five to ten percent.
00:17:22.079 --> 00:17:31.039
That is just to give you a sense of scale, to to align with a 1.5 degree temperature trajectory, which you know we know we've missed already.
00:17:31.279 --> 00:17:35.359
That volume of emissions is half of what you need to get to that level of emissions cut.
00:17:35.440 --> 00:17:36.880
So it's like massive.
00:17:37.200 --> 00:17:40.960
Um, it's like half of the world's most ambitious climate goal.
00:17:41.119 --> 00:17:43.759
And and they were just throwing this out there everywhere.
00:17:43.920 --> 00:17:50.799
Um, and so when I wrote this report, I you know, I sort of went back to an analysis that I did tracing this back to where did this come from?
00:17:50.880 --> 00:17:54.799
Because they talk about like, oh yeah, research shows, you know, analysis shows.
00:17:55.119 --> 00:18:07.519
And it was Boston Consulting Group having a chat, just a casual chat with one of their clients who was like, Oh yeah, I reckon AI has saved me about five to ten percent, you know, of like energy consumption.
00:18:07.759 --> 00:18:16.240
And they were like, Yeah, let's extrapolate that to the entire human project of climate action uh over the next decade.
00:18:17.279 --> 00:18:25.839
Just just shooting the shit with a client, and then all of a sudden uh Google grabs onto this and is like, this is our justification to build all of our gas plants and data centers.
00:18:26.079 --> 00:18:29.039
And I'm not exaggerating, you know, that is very literally what happened.
00:18:29.279 --> 00:18:31.759
And Google is not a small company, right?
00:18:31.920 --> 00:18:35.200
Like they are like highly influential.
00:18:35.359 --> 00:18:41.200
Um, and so that was like the impetus for us being like, okay, well, we need to really dig deeper into all of this stuff, right?
00:18:41.279 --> 00:18:42.559
And find out what's going on.
00:18:42.799 --> 00:18:48.240
What we basically found is first of all, they model up different forms of machine learning.
00:18:48.400 --> 00:19:06.640
So the older stuff like the wind forecasting and the grid management, they refer to that as AI, but they also refer to the massive consumer generative models that people are familiar with, like Chat GPT and Gemini, and they lump it all together and they're like, well, sorry, but you can't have one without the other.
00:19:06.880 --> 00:19:18.720
So all of the lovely climate solutions that Google is funding, um, they're like, Yeah, well, I'm sorry, but if you want that, you also have to let us have, you know, uh these like phenomenally energy intensive models.
00:19:18.880 --> 00:19:24.319
And and in the past, I wish I could describe this chart properly in audio format.
00:19:24.480 --> 00:19:25.279
And I'm gonna try.
00:19:25.440 --> 00:19:31.119
But basically, Google's emission Google's energy consumption has been going up steadily over the past, you know.
00:19:31.200 --> 00:19:33.279
I've got data going back to 2016, right?
00:19:33.440 --> 00:19:44.319
And you go up to 2024, and and it's increasing each year, and then the amount by which it increases slightly accelerates, you know, like it's just sort of ratcheting up by like a terawatt hour, terawatt hour.
00:19:44.559 --> 00:19:50.000
I think every sort of year it was adding about two terawatt hours onto its energy consumption.
00:19:50.480 --> 00:19:54.799
In 2024, that too shifted to 12.
00:19:55.119 --> 00:19:58.640
Um, and then after that, it was I think like 13 or 14.
00:19:58.960 --> 00:20:06.799
And that is like it goes from being not a linear change, like it was accelerating slightly to just being absurd, right?
00:20:06.960 --> 00:20:09.680
Like the acceleration is beyond belief.
00:20:10.000 --> 00:20:32.400
Um, and so Google really ramped up that narrative of like if you want to have nice climate solutions, you have to let us do our generative AI thing, which is you know not literally what they said, but it's by implication, it's like if you regulate us, you won't get any of these lovely climate solutions, kind of holding them hostage in a sense.
00:20:33.039 --> 00:20:39.359
And we debunk that whole idea that you can just sort of smush them together because you can't have one without the other, right?
00:20:39.519 --> 00:20:48.559
You can have the laptop running the wind forecasting thing without also having the you know, 15 trillion gigawatt data center destroying a community's livelihoods.
00:20:48.799 --> 00:21:08.240
We also found, interestingly enough, that even if you ignore that element and you look at just the traditional AI solutions that they talk about, those themselves are not actually that well supported as having like major deep gigaton scale global emissions cuts, like I described, like Google was originally saying.
00:21:08.480 --> 00:21:12.079
They tend to just be like sort of minor efficiency boosts.
00:21:12.240 --> 00:21:14.319
Uh, like you can kind of manage the grid better.
00:21:14.400 --> 00:21:15.599
They're kind of like software solutions.
00:21:15.680 --> 00:21:21.680
It's like me having access to Microsoft Word instead of writing out my reports on paper, right?
00:21:22.000 --> 00:21:29.279
Like it's not a game changer, it's kind of useful, but it's not going to make me, you know, a trillion times more effective, right?
00:21:29.519 --> 00:21:33.759
So um that was the other thing we found is that they tend to use quite weak evidence.
00:21:33.839 --> 00:21:39.200
Like they'll kind of refer to a company statement saying our AI solution is amazing.
00:21:39.279 --> 00:21:47.119
And we thought we were like, okay, well, that's not strong evidence that you can then conclude to say, well, AI is a net benefit for humanity in climate action.
00:21:47.359 --> 00:21:58.400
You know, we've had some success in combating that narrative, like the International Energy Agency is now really careful to distinguish, to carefully distinguish between um the different types of things that they're talking about.
00:21:58.640 --> 00:21:59.599
Google really didn't make that.
00:22:00.160 --> 00:22:02.880
Claim in their latest sustainability report, which was great.
00:22:03.039 --> 00:22:05.119
Um, well, less bad.
00:22:05.440 --> 00:22:10.000
They said a bunch of other stuff, which was a bit annoying, which I criticize in a recent blog post.
00:22:10.480 --> 00:22:21.759
But a couple of days ago, you know, Al Gore is out there, you know, talking to the media, and he's talking about like, oh yeah, you know, like research shows that AI will be a net benefit for humanity.
00:22:21.839 --> 00:22:30.160
And I was like, oh man, you know, we still have a lot of work to do, you know, because we really, it's you know, at the time we're recording this, it's New York Climate Week, right?
00:22:30.240 --> 00:22:32.640
So maybe that's just in the air a bit.
00:22:32.720 --> 00:22:38.559
But um I have been spending a lot of time being very disappointed in people.
00:22:38.720 --> 00:22:45.680
Um, and we've got a lot of work to do to really dispel this, because it's not just the tech companies that say this, right?
00:22:45.759 --> 00:22:54.319
Like there are often a lot of like sort of climate groups that have associations with tech companies or who are generally aligned with them that repeat this as well.
00:22:54.559 --> 00:22:56.480
So a lot to do.
00:22:56.720 --> 00:23:02.559
I'm I'm not surprised by like the disappointment there, you know, but it's just as you're saying, like it's in the air, right?
00:23:02.640 --> 00:23:07.839
It's like what all these people are being told to echo to parrot, and it's also not even being told, right?
00:23:07.920 --> 00:23:20.960
It's just like this is what the narrative is, this just becomes something that people accept, like in the way that you know, so many people started parroting and talking about how AI is apparently gonna like uh threaten humanity in 10 years or something like that, right?
00:23:21.039 --> 00:23:27.920
It just becomes part of the conversation, and then you have these people kind of echoing it, you know, kind of uh continuing to expand it.
00:23:28.079 --> 00:23:39.039
But speaking of those greenwashing claims, you know, one of the people who has been making those and about the data centers, but also about other things, is uh Jensen Wong of NVIDIA, CEO of that company.
00:23:39.200 --> 00:23:52.400
And of course, none of these data centers, certainly the ones for AI, can operate without his chips and you know, the graphics uh processing units that his company makes, they have become absolutely essential to this moment.
00:23:52.880 --> 00:24:04.559
Um, but I, you know, often when we talk about data centers and AI, his company can often be left out in talking about the broader impacts and certainly the climate impacts that come along with that.
00:24:04.720 --> 00:24:15.279
And of course, you know, you looked into that recently because you know it's not just that it gets left out, but the company is also quite uh opaque as far as I understand in what they share and the details that they provide.
00:24:15.440 --> 00:24:26.160
Um, so what do we know about the actual climate impacts of NVIDIA and how that compares to what its leadership actually says about what the company is doing?
00:24:26.799 --> 00:24:34.559
Yeah, I mean, they're one of the most interesting companies in this whole space because they no one qu no one quite knows what to feel or think about them.
00:24:35.039 --> 00:24:49.119
You can't really slot them into a nice category of what we understand to be heavy industrial polluters, because they're kind of like this weird intermediary between the worlds of like the gas plants and the data centers themselves and things like that, right?
00:24:49.200 --> 00:24:53.279
Like they kind of just sit as this weird center of the galaxy.
00:24:53.519 --> 00:24:55.359
They are the richest company on Earth, right?
00:24:55.440 --> 00:24:59.759
Like by market capitalization, by like a good margin, and they have a lot of influence, right?
00:24:59.839 --> 00:25:11.839
Like when I was researching this report, um, which again was just funded by a consortium of climate and corporate accountability groups with Greenpeace East Asia added on as well, because they've been working on NVIDIA issues for a while.
00:25:12.000 --> 00:25:16.000
What we found is that NVIDIA was constantly referred to as the Bank of AI.
00:25:16.160 --> 00:25:16.880
Have you heard this?
00:25:17.039 --> 00:25:17.599
I have, yeah.
00:25:17.759 --> 00:25:21.519
Uh this is like and I was just like, wow, that's fascinating.
00:25:21.599 --> 00:25:25.920
Because in the climate space, uh people have targeted banks for a long time, right?
00:25:26.000 --> 00:25:38.160
Like people have basically said, even though your literal emissions on paper are almost nothing, you are financing and facilitating and earning money from fossil fuels at a scale that is just unbelievable.
00:25:38.400 --> 00:25:49.119
And so I kind of thought, you know, this is this is actually a great thing to start thinking about now because um there's so much focus on the data center developers and the hyperscalers.
00:25:49.359 --> 00:25:55.200
We can expand build on that and expand it outwards in in the context of this specific emissions analysis.
00:25:55.440 --> 00:26:00.720
Nvidia has a lot of scrutiny on around its finances and and its deal making and things like that.
00:26:00.960 --> 00:26:06.240
And what we found, uh, Greenpeace East Asia had done some work on the company's supply chain.
00:26:06.400 --> 00:26:18.640
They buy a lot of these chips that are manufactured in an emissions-intensive way in Southeast Asia, uh, particularly on power grids that have a lot of coal and gas on them, but they're running on electricity.
00:26:18.880 --> 00:26:22.400
A good chunk of the chip manufacturing process is electrical energy.
00:26:22.480 --> 00:26:27.519
So you can, you know, relatively easily fight to decarbonize that element of it.
00:26:27.680 --> 00:26:39.119
Uh, and so Greenpeace, East Asia have been pushing for this, but um, for this one, we went back to 2020 in their sustainability reports just to compare, just to kind of paint the picture of what has happened.
00:26:39.279 --> 00:26:41.200
Uh, and it's quite a phenomenal rise.
00:26:41.279 --> 00:26:49.519
So there's a 725% increase in the emissions of the bits and pieces that they have to buy to manufacture the products, right?
00:26:49.599 --> 00:26:54.880
To make the the hardware that they're selling, the graphical processing units or GPUs.
00:26:55.039 --> 00:27:04.640
I I have another chart in there of uh basically they used to make most of their money from selling video cards, um, and now the ratio is like almost too absurd to even state.
00:27:04.960 --> 00:27:13.599
Like the vast, vast majority of their money comes from selling the bits and pieces for data centers that do AI functions or mostly generative AI functions.
00:27:13.839 --> 00:27:20.880
Um, and so you can pin that increase in emissions on them just manufacturing the hardware to be used in these data centers, right?
00:27:20.960 --> 00:27:26.559
Uh sorry, not manufacturing, but producing, I should say, because they don't actually manufacture or fabricate themselves.
00:27:26.640 --> 00:27:30.720
Um, but they put it all together uh and they package it up and they sell it.
00:27:31.839 --> 00:27:33.920
Today's sponsor is ProtonMail.
00:27:34.079 --> 00:27:38.400
Most email providers don't just offer an email service, they build a business around your data.
00:27:38.559 --> 00:27:42.720
The more they know about you, the more valuable your inbox becomes to advertisers.
00:27:43.039 --> 00:27:45.359
Proton takes a completely different approach.
00:27:45.519 --> 00:27:48.319
It's funded by subscribers, not by advertisers.
00:27:48.400 --> 00:27:50.559
So you're the customer, not the product.
00:27:50.720 --> 00:27:59.279
That means no ads following you around the web, no profiling based on your emails, and no business model built on collecting more of your personal information.
00:27:59.440 --> 00:28:02.640
And making the switch is much easier than most people expect.
00:28:02.880 --> 00:28:08.720
EasySwitch lets you move your emails, contacts, and calendars in just a few clicks, so you don't have to start over.
00:28:08.960 --> 00:28:14.720
ProtonMail is based in Switzerland, owned by a nonprofit, and every email is end-to-end encrypted by default.
00:28:14.880 --> 00:28:16.640
It's built around a simple idea.
00:28:16.880 --> 00:28:20.480
Your inbox should belong to you, not the company providing it.
00:28:20.640 --> 00:28:24.079
When I left Gmail, it's the option that I chose for my email service.
00:28:24.240 --> 00:28:28.079
As a Tech Won't Save Us listener, you can get your first month for just one dollar.
00:28:28.240 --> 00:28:31.119
You'll find the exclusive link in the show notes to do that.
00:28:31.279 --> 00:28:33.200
Start now for just one dollar.
00:28:34.640 --> 00:28:48.400
Does that make it easier for them to kind of hide their emissions or kind of obscure their emissions, where you know they often get TSMC or one of these other companies to actually do the manufacturing piece and then they kind of claim it's off their books or something?
00:28:48.480 --> 00:28:50.240
Like how are you seeing that?
00:28:50.559 --> 00:28:51.359
Yeah, absolutely.
00:28:51.440 --> 00:28:53.920
And they and they it is technically on their books, right?
00:28:54.000 --> 00:29:08.319
Like they report uh what's known um in sustainability nerd speak as scope three, which is basically just the emissions outside the company when you buy something or when you sell it, what happens outside of the boundaries of what you're sort of personally responsible responsible for.
00:29:08.480 --> 00:29:18.880
They do report it, but one thing that's missing, you know, really badly from their report is just like where those chips come from and the individual details of how they're dealing with those suppliers.
00:29:19.039 --> 00:29:21.759
Um, one company that does this really well is Apple.
00:29:21.839 --> 00:29:23.839
Uh, Apple has so much detail in there.
00:29:24.079 --> 00:29:31.039
I mean, I have a lot of gripes with how they, you know, do their emissions and who knows how the AI stuff is going to change that for them.
00:29:31.359 --> 00:29:41.599
But they do report their supplier engagement really, really well and with enough detail that you can scrutinize it and go, well, you're leaving out, you've got this massive gap over here, or you're doing really well over here.
00:29:41.839 --> 00:29:50.880
Nvidia just they give you a headline number, which is the number that we talk about in the report, but they don't give you enough detail to scrutinize it, which sucks because they have so much power.
00:29:51.039 --> 00:29:53.519
They should absolutely be disclosing more information on that.
00:29:53.680 --> 00:30:03.279
And then the other side of the story, in terms of their indirect emissions, or like a kind of you know what they would describe as outside of their responsibility, uh, is what happens after they sell it.
00:30:03.519 --> 00:30:12.240
They have like a big, you know, graphical processing unit that's full of the chips that they've bought and that that's been manufactured, and they sell it to a company and it gets put into a data center and used.
00:30:12.480 --> 00:30:17.599
And companies like Intel and AMD, they all report what happens in this process, right?
00:30:17.680 --> 00:30:22.640
Like after you sell it, um, Microsoft actually reports all of its like product use emissions.
00:30:22.720 --> 00:30:41.039
So when somebody buys like an Xbox or Surface Pro, it's actually a huge chunk of their total emissions, is like the grid emissions of just plugging something into the War Socket in your home, and then a coal plant or a or a gas plant spins up a little bit more to generate to run that thing, including renewables.
00:30:41.200 --> 00:30:49.680
So, again, I have lots to criticize about Microsoft, but they have infinite money, you know, they can afford to disclose this stuff and calculate it because it doesn't take a lot of effort.
00:30:49.839 --> 00:30:52.240
And so we thought that was really shocking for NVIDIA.
00:30:52.559 --> 00:30:57.599
Just to give you an example of how weird it is that they don't disclose this number.
00:30:57.759 --> 00:31:01.920
There was a shareholder resolution brought forward basically saying, like, what are you doing?
00:31:02.000 --> 00:31:03.839
You know, please just share this number.
00:31:03.920 --> 00:31:10.000
Like, that's just like basic information we need to know about who you're selling your chips and hardware to.
00:31:10.240 --> 00:31:14.880
And the oil fund voted in favor of the shareholder proposal.
00:31:14.960 --> 00:31:18.720
That's Norway's oil fund, the largest sovereign wealth fund on the planet.
00:31:18.960 --> 00:31:27.519
Nvidia is the oil fund's largest single investment, and it's like 1.6% of the total oil funds investments, which is pretty substantial, right?
00:31:27.599 --> 00:31:30.960
Like everything else is very, very diversified across a lot of sectors.
00:31:31.200 --> 00:31:35.359
But it a lot of stock is held for the Norwegian citizens in Nvidia, right?
00:31:35.440 --> 00:31:43.680
Like I think I personally, you know, probably you could attribute some of that to my, you know, like, because it's all Norwegian people have a part of this.
00:31:43.839 --> 00:32:03.680
And even the oil fund, whose CEO has the worst case of AI psychosis I've ever seen in any executive ever, there is a LinkedIn video of him cutting a cake for his birthday, and it's an agent, it's like a robot-shaped cake that he's cutting into because he's just so obsessed with AI.
00:32:03.920 --> 00:32:10.640
Even the oil fund was like, Nvidia, you should really be sharing your uh product use emissions, you know.
00:32:10.880 --> 00:32:19.839
So if the oil fund, you know, which is a fund of fossil fuels with an AI-obsessed CEO, is saying, Why aren't you sharing this?
00:32:20.000 --> 00:32:22.000
This is like the absolute bare minimum.
00:32:22.240 --> 00:32:26.000
It's it's why is not the why is the richest company in the world not sharing this?
00:32:26.160 --> 00:32:28.160
Like it's it's a really shocking thing that we found.
00:32:28.240 --> 00:32:30.000
And so we estimated it ourselves.
00:32:30.240 --> 00:32:36.000
Epoch AI uh has estimates of the total power consumption of the chips they've sold.
00:32:36.240 --> 00:32:52.000
We did the numbers, we ran it against some like grid emissions factors, and we found that Nvidia's product use emissions are probably around about the same product use emissions of the Russian-owned coal companies, a state-owned coal company.
00:32:52.160 --> 00:32:56.799
Um, it's like about sort of, it could be up to about like 20 to 25 megatons.
00:32:57.279 --> 00:33:04.480
A very, very conservative estimate because we don't include all of the cooling and the overheads and everything else in a data center.
00:33:04.720 --> 00:33:12.880
We use a conservative utilization factor, which is basically of all the power it could consume, how much do you think it will, which is like how hard they run the chips.
00:33:13.119 --> 00:33:21.519
And so all these numbers we were really careful to take a very conservative view, and it still comes out just like massive, you know, like really, really huge.
00:33:21.680 --> 00:33:52.559
Um, and so we follow that up by talking about um a deal that got announced, you know, in the sort of late stages of writing this report, with this one project, this phenomenal project, uh, which I think I mentioned earlier, which is the Ports Pike Gas Generation Facility in Ohio, which is a data center, um, has a 9.2 gigawatt uh gas plant proposed, which on the list of all proposed gas plants is number two uh, you know, planet wide.
00:33:52.720 --> 00:33:54.079
Um that's the second biggest.
00:33:54.240 --> 00:34:00.960
Number one is it also a data center in the US, um, and is only a couple of you know hundreds of megawatts more.
00:34:01.279 --> 00:34:08.800
If it were to be built, um it would be twice the size of the next biggest operating gas plant, which is a Saudi Arabian facility.
00:34:08.960 --> 00:34:19.280
And then if it were to be built, it would be half the size of the largest power station ever constructed, which is the Three Gorgeous Dam in uh hydro plant in China.
00:34:19.599 --> 00:34:31.119
Nvidia trumpeted very loudly, look at us, you know, we're the exclusive supplier of chips to this project with great pride, and you know, they really blasted it out anywhere, everywhere.
00:34:31.360 --> 00:34:42.800
This is the soft bank energy project, which is, you know, uh like there's a lot of I know there's a lot of like, you know, very complicated financial relationships going on there that I don't fully understand.
00:34:43.039 --> 00:34:55.760
But what I do get is that Nvidia saw this like one of the most significant fossil fuel projects ever proposed of any kind, uh, and was just like, yes, we want to be the sole supplier of chips to this.
00:34:56.079 --> 00:35:11.119
And so that you get an idea of why NVIDIA would be like, no, I'm sorry, we're not disclosing our um scope three you know user emissions, because even if investors as big as the oil find are saying we want to know this information, NVIDIA doesn't want them to know it, obviously, right?
00:35:11.199 --> 00:35:14.320
Because they're just they're not sharing this information that they could easily could.
00:35:14.719 --> 00:35:16.559
Yeah, it's it's shocking to hear, right?
00:35:16.719 --> 00:35:26.719
Like, especially where I feel like that dimension of the conversation doesn't so often enter into the discussions about the impacts of data centers and and that sort of a thing.
00:35:26.880 --> 00:35:55.840
I I wonder, you know, having done this research and looked into the emissions from NVIDIA, how do you think that complicates or or kind of adds to the broader discussion about data centers and AI and the climate impacts that we're that we're seeing from it when we start to look not just at what the data centers exist, the power, but the supply chain that then feeds into them that is essential for powering all this AI and for you know the justification for the data center in the first place.
00:35:56.400 --> 00:36:11.280
Yeah, I think it brings it starts to bring the conversation into more of a traditional climate and fossil fuel discussion, which is that you've kind of got this webbed network of different players who all make money at different points in the supply chain for fossil fuels.
00:36:11.519 --> 00:36:27.360
I'm sure you would have seen like those memes that sort of go around where it's like the numbers always change, but it's always something along the lines of like a hundred companies are responsible for 80% of emissions, and they're kind of mangling what is actually quite a meaningful report, right?
00:36:27.440 --> 00:36:36.400
Which is this annual report that looks at not the companies that burn the fossil fuels, but the ones that produce the fossil fuels, right?
00:36:36.480 --> 00:36:45.199
So, and that percentage that you hear, you know, 100 companies, blah blah blah, it's referring to that, which is the first step in unlocking fossil fuels from underground.
00:36:45.519 --> 00:36:49.119
And those companies actually have some responsibility, right?
00:36:49.280 --> 00:36:55.679
Because they're making obscene amounts of money by being the ones to dig it up and process it and supply it to the world.
00:36:55.920 --> 00:37:06.159
And by being like such eager and excessive oversuppliers of fossil fuels, that encourages more people to burn fossil fuels and use them.
00:37:06.400 --> 00:37:11.760
And I think that Nvidia are really like one of the first, you know, AI majors or something.
00:37:11.840 --> 00:37:19.679
I don't know where you can think of a better term for that, but like they are one of the first like absurd oversuppliers, right?
00:37:19.760 --> 00:37:37.280
Um, because they just want to just pump out this product everywhere in the world, do as many deals as possible, make as many of these like complicated, you know, cross like webbing like finance deals as possible, just to cement this in as something that people rely on and use way, way, way too much.
00:37:37.440 --> 00:37:40.400
Um, and obviously that intersects with the fossil fuel industry.
00:37:40.559 --> 00:37:48.719
And I think that we can start to understand the AI environmental and climate story in a really similar way to what how we understand fossil fuels, right?
00:37:48.800 --> 00:38:00.639
Which is that you kind of just have this like stack of terrible companies, each, you know, taking their cut in some awful way until it ends up with it being burned, you know, for whatever reason in whatever machine.
00:38:00.800 --> 00:38:10.960
Uh, you can think of the same thing with just like extremely energy inefficient um computing hardware being being used for a process that is itself very, very inefficient.
00:38:11.119 --> 00:38:12.000
So yeah.
00:38:12.400 --> 00:38:29.360
And when we look at the hardware itself, because I know that you touched on this in the report as well, like I know that NVIDIA and and Jensen Wong like the claim that the hardware is getting so much more efficient and and those sorts of things, but what do we actually see in the hardware itself, its energy use?
00:38:29.519 --> 00:38:34.159
And obviously, it is you know helping to drive some of these wider things, right?
00:38:34.239 --> 00:38:46.639
Because if they if these are the chips that you're buying, then obviously that requires a certain amount of um of energy in order to power them, and you know, they are going to be what kind of AI is designed around or designed to use and that sort of thing.
00:38:46.800 --> 00:38:48.000
What do we see in the hardware?
00:38:48.159 --> 00:38:51.119
Like, how is that kind of fueling this whole picture, too?
00:38:52.159 --> 00:39:05.280
There's a great line uh from the sustainability guy, and it I I can't remember how many zeros were in this number, uh, but he says that something along the lines of like we've had like a hundred thousand times or ten thousand times efficiency increase since the 2020s.
00:39:05.440 --> 00:39:06.800
And I'm sure it's true, right?
00:39:06.880 --> 00:39:10.159
Like, I wouldn't be surprised if that's you know roughly, roughly correct.
00:39:10.400 --> 00:39:19.280
But it's funny to then compare it to what we see in terms of their emissions impact, which has increased, you know, several hundred several hundred or you know, even thousands of times potentially.
00:39:19.440 --> 00:39:27.920
I think that's really important because the efficiency narratives work really well for, you know, like let's just say Climate Week New York crowds, right?
00:39:28.000 --> 00:39:30.239
Where people are like, oh great, you know, you're becoming more efficient.
00:39:30.320 --> 00:39:32.159
That's what we want, that's what we want to hear.
00:39:32.400 --> 00:39:42.559
But uh people often talk about this thing called Jevon's paradox, which is this idea that you know, as you make something more efficient, it just incentivizes more headroom for more to be used.
00:39:42.800 --> 00:39:44.559
But I don't really use that term, right?
00:39:44.639 --> 00:39:46.880
Because it's not a paradox, it's more of a choice.
00:39:47.039 --> 00:39:53.760
It's more of like, you know, we made something more efficient and then we just did the worst possible next step, which is that we use more of it.
00:39:54.000 --> 00:39:56.159
We see lots of good efficiency, right?
00:39:56.239 --> 00:40:03.840
Like uh as uh household fridges became more efficient, we didn't just buy a second fridge because you only really need so much fridge space.
00:40:04.000 --> 00:40:09.119
I mean, I guess some people probably did buy a second fridge, but not enough to undo the efficiency gains, right?
00:40:09.360 --> 00:40:20.400
When it comes to these chips, uh what we found in this report is that their most efficient chip has also been sold at the greatest volume and the product size, right?
00:40:20.480 --> 00:40:23.599
They're cramming eight cores into this like one thing, right?
00:40:23.760 --> 00:40:25.360
And they basically see it.
00:40:25.599 --> 00:40:32.559
Sometimes they talk about it in the context of sustainability, so they're like, yeah, this is sustainable innovation, you know, our most efficient chip, blah blah blah.
00:40:32.719 --> 00:40:49.599
Uh, and then other times, when I guess they think that people like me aren't really listening, uh, they start talking about how explicitly the purpose of that is to be able to uh have customers fit more hardware and get more um AI out of the one, you know, one product purchase.
00:40:49.840 --> 00:40:57.119
What we see, of course, then is the effect on the real world, which is that the emissions and energy impacts of these facilities go up.
00:40:57.280 --> 00:41:00.000
Um, the more efficient they get, the worse the problem gets.
00:41:00.239 --> 00:41:07.039
That is just the nature of the desperation and the psychosis and the and the sort of impatience behind all of this infrastructure build out.
00:41:07.519 --> 00:41:19.199
It makes me think as well about the major companies, you know, as we were talking about earlier, where they will discuss how they have this new kind of way of building a data center or a cooling system in a data center that makes it efficient or or whatnot.
00:41:19.360 --> 00:41:24.800
But they certainly don't want you to look at how they're building so many more data centers to make that completely irrelevant, right?
00:41:24.880 --> 00:41:30.960
Like, you know, the the single data center is more efficient or you know, more uh less energy intensive or something like that.
00:41:31.039 --> 00:41:35.440
But they're building so many data centers alongside it that it's completely irrelevant.
00:41:35.519 --> 00:42:04.880
Um, but you know, speaking about emissions, speaking about gas as we have, uh, Australia also enters into this picture, you know, obviously a country that is very dependent on its uh fossil fuel industry and and that has really been shaped and kind of captured by the fossil fuel industry in many ways for a long time, but still wants to appear as though it is kind of acting on climate change and is not a climate villain uh in the way that we see some other countries around the world.
00:42:05.039 --> 00:42:08.320
You know, data centers have become a huge issue in the United States, right?
00:42:08.480 --> 00:42:10.320
And have been for a little while now.
00:42:10.480 --> 00:42:14.880
Um, but we are seeing more and more countries start to experience something similar, right?
00:42:14.960 --> 00:42:24.559
In Canada, where I am, uh, you know, in the past kind of six to nine months, we have really seen not just the proposals for data centers, but the opposition to data centers really take off.
00:42:24.719 --> 00:42:36.400
And so I wonder what has been happening down in Australia and how its existing kind of dominance of the fossil fuel industry has played into the conversation happening down there.
00:42:37.280 --> 00:42:37.920
Yeah.
00:42:38.159 --> 00:42:50.880
Uh I uh did a recent report for Greenpeace Australia Pacific uh on this, and we kind of covered two big areas, um, which I think still summarize it quite nicely, and I can maybe quickly talk about a third as well.
00:42:51.039 --> 00:42:56.000
The first is basically just what happens when you plug a lot of big data centers into the power grid in Australia.
00:42:56.320 --> 00:43:00.719
There's been not anywhere near enough renewable energy getting built, which is so strange.
00:43:00.800 --> 00:43:04.880
You wouldn't know it by looking at the way you know renewables and energy are covered in Australia.
00:43:04.960 --> 00:43:11.519
It kind of there's a lot of uh I would say premature uh victory, you know, sort of being celebrated.
00:43:11.760 --> 00:43:16.079
Emissions are falling, but they're not falling anywhere near as fast as was predicted.
00:43:16.239 --> 00:43:21.199
And that's a problem because there's just not enough renewables being built for the climate and energy targets they've set.
00:43:21.360 --> 00:43:24.320
And so then you add in data centers into the mix, right?
00:43:24.480 --> 00:43:32.559
Where the the target, the goal where the renewables we're trying to get to is just move further away because data centers just increase their future demand.
00:43:32.800 --> 00:43:39.760
So you have to build extra renewables just to be in the same spot that you were before, let alone uh actually, you know, advance.
00:43:40.559 --> 00:43:50.880
And so there's this whole narrative of like data centers are helping the energy transition by incentivizing new renewables, but they're adding a lot of new demand and only adding a tiny sliver of uh renew renewables.
00:43:51.519 --> 00:43:54.000
And so that the net result of that is just more emissions.
00:43:54.159 --> 00:43:59.760
Like that is just mathematically what happens, and I have been so mad to see people.
00:44:00.159 --> 00:44:05.920
People fall for this line of like from the tech industry being like, oh no, no, we're helping the energy transition, we're helping climate change.
00:44:06.079 --> 00:44:09.599
But just mathematically, there is no way they're doing that.
00:44:09.760 --> 00:44:19.679
Uh unless they were to build like double the amount of renewables as the data centers that they build, which they just they would never do because they aggressively oppose that even you know the basic concept.
00:44:19.920 --> 00:44:21.119
So that's one problem.
00:44:21.360 --> 00:44:25.119
The second problem here is that the gas industry in Australia has been freaking out.
00:44:25.280 --> 00:44:41.920
Uh they don't really think that they can keep selling gas to Asia, which has been really fucked over by the conflicts, like all of the you know, geopolitical issues and conflicts that have arisen in the past few years, not just not just the Iran homoose crisis, but um over the past half decade.
00:44:42.159 --> 00:44:46.159
Gas just keeps fluctuating so wildly and people just get more and more angry.
00:44:46.239 --> 00:44:49.360
And so it's it's destroying a lot of demand for gas in Asia.
00:44:49.519 --> 00:44:52.639
And all these Australian producers are like, shit, who's gonna buy our gas?
00:44:52.800 --> 00:45:00.880
You know, they've they have so many planned gas fields, and so data centers come along and they're like, oh my god, yes, this is perfect, you know.
00:45:01.039 --> 00:45:16.159
Um, and the narrative that has been constantly going around is actually specifically data centers for training AI being built in like the northern and sort of west northwest Australian regions where there's a lot of like planned gas fields.
00:45:16.320 --> 00:45:25.679
So you just plonk a data center there and you run it entirely on gas, like you build gas power stations on site, like literally next to the gas field.
00:45:26.000 --> 00:45:36.320
Like in the report, um, I muck around with some of the calculations for the climate impacts of this, and it's like you know, one just one of those fields, the entire Northern Territory.
00:45:36.400 --> 00:45:42.480
It would be like more emissions in the entire Northern Territory, all of the pre-existing industry and fossil fuel industry there.
00:45:42.639 --> 00:45:45.440
Because they're just like, Yeah, yeah, we'll do five gigawatts, who cares?
00:45:45.519 --> 00:45:46.000
You know?
00:45:46.400 --> 00:45:52.079
So they're kind of throwing around these ideas, and they sound ridiculous, right?
00:45:52.320 --> 00:45:59.039
But there are some very, very serious American and other um gas companies who are getting involved now.
00:45:59.280 --> 00:46:05.360
The plans are actually solidifying, and they're becoming more and more likely as every day goes by.
00:46:05.599 --> 00:46:06.960
So that's the second thing, right?
00:46:07.039 --> 00:46:13.760
Like it's just the desperation of the gas industry is driving a lot of these data center, fossil fuel data center projects.
00:46:14.000 --> 00:46:28.000
And then the third thing that has kind of emerged more since we wrote this report in May has been the government basically saying, chill out, we're very responsible, we're not just doing this like willy-nilly, we're gonna do very responsible regulations for data centers.
00:46:28.159 --> 00:46:38.079
Um, and there are lots of people who are like experts on like, you know, community impacts and water impacts, who've basically called these um proposed government regulations out as insufficient.
00:46:38.239 --> 00:46:42.960
But my area of expertise is climate and energy, and they're so bad.
00:46:43.199 --> 00:46:59.360
What I basically found in a recent blog post is that the the data center lobby basically demanded that the independent statutory body that that provides advice to the government on this have so the government said we want to have 100% renewables for all new data centers.
00:46:59.760 --> 00:47:20.880
And then this body said, Yeah, you can do that, uh, but what we suggest is that you do it through this like offsetting scheme, which is basically uh you can buy a certificate from a pre-existing renewable energy facility, just zoom out, right, and imagine this situation.
00:47:20.960 --> 00:47:29.840
So you've got a data center that just gets plonked into Australia, and then they're like they're buying these certificates from an existing wind farm, so they don't add any new renewable energy.
00:47:30.000 --> 00:47:31.679
So they're just drawing from the grid.
00:47:31.840 --> 00:47:43.440
So you have a situation where emissions on the grid are going up, but if you were to look at the regulation and you look at the climate reports of all the companies and all the government statements, you would say that they're all 100% powered by renewables.
00:47:43.519 --> 00:47:44.320
What's the problem?
00:47:44.480 --> 00:47:47.599
I don't see what any issue is here, right?
00:47:47.840 --> 00:47:59.840
Um, and then they propose this thing called a glide path, which is like it eventually does require new renewable projects, but even then you can't say that that new renewable project was from the data center company.
00:48:00.000 --> 00:48:04.880
It just could just be what happens anyway, because you know, new projects are getting built, so it's the same problem.
00:48:05.119 --> 00:48:13.840
So Island had this, you know, as you know, like this massive moratorium on data centers, and one of the reasons was to have time and space to think about regulation.
00:48:14.320 --> 00:48:25.679
But then they proposed this glide path concept where you basically just have like eight years of just pumping emissions into the atmosphere, and then they're like, Yeah, yeah, yeah, we'll we'll we'll deal with it in eight years, you know.
00:48:25.760 --> 00:48:29.840
Like uh I'm sure we'll have new renewables by hand, so don't worry too much about it.
00:48:30.079 --> 00:48:58.960
And it's basically just saying just don't worry about the next sort of half decade to one decade, because it's so urgent that we build these data centers and capture this opportunity that they're kind of formally saying, fuck it, we don't care about the emissions, we just want this to happen, and we don't care about the harm that will occur from the power system that they're drawing from or the new gas plants that get built.
00:48:59.119 --> 00:49:01.280
Uh, and that is really shocking to me, right?
00:49:01.360 --> 00:49:15.440
Like I found that to be such a a failure of basic considerations on this question, but it didn't really draw a lot of outrage or criticism because I think people find it a bit confusing and a bit complicated, right?
00:49:15.519 --> 00:49:20.960
They're like, oh yeah, but no, they said that with that it was like 100% renewables, you know, so I don't understand what the problem is.
00:49:21.199 --> 00:49:24.719
Um, so I that's one thing that I really fail to crack, right?
00:49:24.800 --> 00:49:45.920
Like it's just I think it's like a an ongoing project, but it's really, really important because we're now in the phase where governments are like listening to the angry cries of the public, and then like, oh fuck, okay, I've got now I've got to like propose something, and they and and what they might end up proposing is something that looks like regulation, but isn't really regulation.
00:49:46.079 --> 00:49:50.559
Um, that is a massive risk at the moment, um, like faux faux regulation.
00:49:50.960 --> 00:49:54.320
And you can see this occurring in many different parts of the world.
00:49:54.480 --> 00:49:58.400
There was a US state just a few days ago, oh my gosh, I wish I could remember what it was.
00:49:58.559 --> 00:50:04.320
I think it was Pennsylvania, where they proposed a ban on gas power for data for data centers.
00:50:04.480 --> 00:50:11.519
Um, and but then you you look a little bit deeper into the wording, and it's actually a ban on 100% gas.
00:50:11.760 --> 00:50:16.000
So it could be 99.9% with one solar panel on the roof, you know.
00:50:16.320 --> 00:50:17.039
That's so wild.
00:50:17.119 --> 00:50:21.920
Every every data center in Pennsylvania now just has the the one uh single solar panel up there.
00:50:22.000 --> 00:50:23.840
It's like, hey, we're getting some, you know.
00:50:24.159 --> 00:50:25.840
Yeah, sorry, I need to check that state.
00:50:25.920 --> 00:50:36.719
But yeah, it's it's a it's a US state, and it's just you know, a nice example of where uh a time poor journalist will write that up as a win for environmental regulation of data centers.
00:50:36.960 --> 00:50:41.920
But then you spend, you know, half an hour just reading the document and you're like, that's not a win.
00:50:42.239 --> 00:50:46.880
I I wanted to ask you, you know, what you've been telling us there is obviously very concerning, right?
00:50:47.199 --> 00:50:51.519
Unfortunately, not surprising, but it's certainly not what you would want to see.
00:50:51.599 --> 00:51:00.960
And it's no surprise that people in Australia are starting to really push back on, you know, what is planned here, what the industry is planning and how the government is enabling it.
00:51:01.199 --> 00:51:16.159
Can you talk to me a bit about how uh, you know, the Prime Minister Anthony Albanese, who's with the Labour Party, is handling this and is trying to like thread the needle between, on the one hand, accepting all this investment, but also trying to make it palatable to the public.
00:51:16.400 --> 00:51:40.639
And also, on the other hand, how we're seeing kind of the lobby groups for the tech industry and the data center industry and the fossil fuel industry in Australia come together in a really significant way that is certainly we have seen in other places, you know, we've seen in the United States, but feels very notable to see the degree to which these two industries are coming together and seeing common cause on gas in particular.
00:51:41.039 --> 00:51:55.360
Yeah, I'll start with the second question because just yesterday the the head of the data center lobby in Australia uh attended uh a conference, a news corp conference that was, you know, entirely funded by the fossil fuel industry and was basically a fossil fuel industry conference, right?
00:51:55.440 --> 00:52:02.800
Like they call it energy, but it's it was all it was just like a litany of all the biggest fossil fuel companies just talking about um how great fossil fuels are.
00:52:02.880 --> 00:52:23.920
Um and of course, you know, the tech companies are there and the data centers are there, and they're all you know basically discussing, in their very carefully worded language, about reliability and the need for stable 24-7 power, which has always been their code words for just burning the dirtiest, crappiest fuel that you can for electrical energy, right?
00:52:24.239 --> 00:52:34.239
And so it went in the report, in the Greenpeace report, we also detail how uh the head of that data centers group actually went to the literal fossil fuel industry conference in Australia.
00:52:34.400 --> 00:52:37.920
It was on a panel sponsored by Chevron, our friends at Chevron.
00:52:38.159 --> 00:52:41.199
Um, and you know, they're basically talking about the same thing, right?
00:52:41.360 --> 00:52:51.920
Where what they'll do is they'll kind of throw out a few words in favor of renewables, but then they start talking in this special code about reliability and and you know stable 24-7 power.
00:52:52.079 --> 00:52:58.800
And it's the way it's the way that they talk to each other about fossil fuels, uh, and the media outlets will just be like, oh great, okay.
00:52:58.960 --> 00:53:06.639
They you know, very literal-minded journalists will just write it up as if they're talking about like they care about reliability when that is not really how the grid works, right?
00:53:06.719 --> 00:53:09.679
Like that is not how you, you know, the reliability of a grid occurs.
00:53:09.840 --> 00:53:11.199
So that's really bad.
00:53:11.280 --> 00:53:19.039
Um, and then Albanese in particular has been a really fascinating, you know, person to watch with this data centers issue.
00:53:19.119 --> 00:53:21.119
Uh, I think it really frustrates him.
00:53:21.199 --> 00:53:29.599
He doesn't Albanese is a is a politician who uh is only happy when he's in this special little zone of like the issues that he cares about and thinks about.
00:53:29.840 --> 00:53:35.679
Anything else, tech, data centers, climate change, energy, he hates all that stuff.
00:53:35.760 --> 00:53:37.360
And you can see it in his face, right?
00:53:37.440 --> 00:53:41.280
Like he looks so strained, he's just not interested.
00:53:41.440 --> 00:53:46.800
Uh I don't think he even cares that much about fossil fuels because he's just not interested in it.
00:53:47.039 --> 00:53:54.400
It's just kind of like, you know, he needs to go along because there's such a powerful lobby in Australia, but like don't have to make me deal with this too much.
00:53:54.480 --> 00:53:57.119
I have my other issues that I prefer to be talking about.
00:53:57.280 --> 00:53:57.760
Yeah.
00:53:58.079 --> 00:53:58.639
Absolutely.
00:53:58.719 --> 00:54:01.360
And it goes the same way with it with most issues, right?
00:54:01.519 --> 00:54:11.280
Where you know there'll be a public outcry, something will force the government to pay attention to something, and then they'll be like, okay, we're gonna regulate it, we're gonna do something about this, don't worry, we're in charge.
00:54:11.440 --> 00:54:18.559
And then the and then the lot like the corporate lobby group will just come along and very, very gently suggest something, and the government is like, ah, yep, okay, that's it.
00:54:18.639 --> 00:54:19.920
Yep, we're doing that.
00:54:20.880 --> 00:54:22.639
Just instantaneously, you know.
00:54:22.719 --> 00:54:30.239
Um, and you can tell by the way that the lobby groups respond because they're always like, This is a very sober and pragmatic and realistic set of regulations.
00:54:30.400 --> 00:54:35.119
You know, we agree with the very adult, serious uh, you know, policies proposed here.
00:54:35.280 --> 00:54:38.159
And then it's like, okay, well, yeah, like that's their victory cry.
00:54:38.320 --> 00:54:42.960
That's the victory cry of the corporate lobby group is uh look how pragmatic this is.
00:54:43.119 --> 00:54:45.519
Um, and so that's exactly what's happened, right?
00:54:45.599 --> 00:54:47.119
Um they're like, yeah, this is really good.
00:54:47.199 --> 00:54:48.719
You know, we agree with all this.
00:54:49.440 --> 00:55:06.880
In particular, they don't want to be made to have the responsibility of the impacts on the power grid, so not just emissions, but the costs that that gets imposed, um, the new transmission lines and substations and all that extra stuff that you need to build to accommodate at the scale they want to.
00:55:07.039 --> 00:55:09.440
You mentioned before the comparison to the US.
00:55:09.599 --> 00:55:23.679
Um, I did a calculation a couple of days ago that basically just, you know, it the percentage of data centers on the US grid, which is, you know, comfortably number one in the world, will probably be about the same as Australia.
00:55:23.760 --> 00:55:35.199
If, you know, it kind of goes by like the slightly higher end of projections for Australia, and you go by the middle scenario in the US, it will be the same by 2030 or sort of like in the early 2030s.
00:55:35.360 --> 00:55:38.239
And so I don't think people realize that, you know, like Australia.
00:55:38.320 --> 00:55:43.599
There was another report, you know, a few days ago, Australia's like number three in the world for I think they measured it by investment.
00:55:43.920 --> 00:55:54.480
So it doesn't have the biggest projects, but as a proportion of the country, it's like a disproportionate amount of data center capacity occurring in Australia, like one of the biggest in the world for sure.
00:55:54.719 --> 00:56:17.760
And so all of these dynamics, so the fossil fuel industry getting super excited, the tech industry carefully but clearly getting involved with the fossil fuel industry, and then the government trying to like get the public response of like thank you for you know regulating these scary, you know, like this scary new industry, but not actually regulating them properly?
00:56:17.920 --> 00:56:25.360
Um, that is a really nice set of dynamics that I think other parts of the world can look at and be like, we do not want to do what Australia is doing.
00:56:25.679 --> 00:56:26.719
Definitely, definitely.
00:56:26.960 --> 00:56:42.960
And I wonder, you know, once we see that kind of reliance on gas, but also the significant expansion in data centers planned in Australia and happening in Australia, what does that mean for the country's climate targets or, you know, kind of any indication that that they might get there?
00:56:43.119 --> 00:56:57.360
And is the growing public opposition in Australia starting to challenge some of the ambitions of the industry or or force governments, whether on the federal level or the state level, to actually start to maybe do something about this?
00:56:58.320 --> 00:56:59.760
Oh my gosh, that's a great question.
00:56:59.920 --> 00:57:06.000
Well, on the climate targets element, I I maintain a sort of tracking process basically for Australia's emissions.
00:57:06.079 --> 00:57:08.719
I've been doing it since, you know, before I before I left.
00:57:08.880 --> 00:57:14.159
And what I do is I track every projection of future emissions against what actually happens, right?
00:57:14.239 --> 00:57:17.199
Like the measured uh emissions of the country.
00:57:17.440 --> 00:57:20.239
And for for a while, Australia was really overperforming.
00:57:20.320 --> 00:57:21.199
Like, do you know what I mean?
00:57:21.280 --> 00:57:28.639
Like that, like I think it was a mix of it was a mix of like very pessimistic projections and underestimated renewable energy success.
00:57:28.880 --> 00:57:36.400
Um, but for the past sort of three or four years, and I don't think this is well recognized like within the country, Australia has been underperforming, right?
00:57:36.559 --> 00:57:51.360
So it's it's even though it has like this amazing in particular with solar growth, that is still not enough to deal with the fact that like not just current but future demand has been growing way, way more than anybody expected.
00:57:51.679 --> 00:57:54.159
And renewables haven't been getting built enough.
00:57:54.320 --> 00:58:25.840
Um, and so the net result of that is that Australia is not only projected to not hit its 2030 um and 2035 climate targets, Australia is getting further, like the gap in between those lines is getting bigger, like a little bit bigger and bigger each year, um, as each year's projection accounts for more and more load growth in the power sector from data centers, along with like a few other factors, like uh there's more uh demand from uh cooling in summer because climate impacts are worse than expected.
00:58:26.079 --> 00:58:37.599
So there's sort of a sort of collection of things, but this is one of the things that's big enough for the government departments themselves to mention in these annual updates, where they're like, uh yeah, it's like data centers.
00:58:37.760 --> 00:58:42.480
Like, you know, for them to explicitly say it and admit it in the reports is a big deal, I think.
00:58:42.639 --> 00:59:09.360
There are elements of the climate movement and like energy and sort of pro-electrification spaces in Australia and I guess other parts of the world too, that are either dismissive or in favor of um tech growth um and data centers, particularly in this narrative of like, well, we can just power it all cleanly, you know, and be like an exporter of AI, you know, services to the world because it's all running on our clean solar power.
00:59:09.599 --> 00:59:13.920
And it's a nice idea, but it's just not what's playing out at all.
00:59:14.159 --> 00:59:22.400
The companies aren't building renewables, the government isn't incentivizing renewables, Australia is still the number one per capita coal power user on the planet.
00:59:22.639 --> 00:59:32.000
Australia's two biggest data center states, New South Wales and Victoria, have double the emissions intensity of the three states in the US with the most data center growth.
00:59:32.159 --> 00:59:46.960
I don't think people realize um how much worse it is to build a data center in Australia for emissions um than it is to do it in like these like already long-suffering American states where communities are so pissed off and the power grids are really dirty.
00:59:47.119 --> 00:59:48.480
Australia is still worse.
00:59:48.639 --> 01:00:00.800
Um and so, you know, this dream of like we'll be like an exporter of green AI to the world uh is really greats on me because I think that it's quite an irresponsible thing to suggest.
01:00:01.440 --> 01:00:03.760
It's okay though, because gas is green, right?
01:00:03.840 --> 01:00:06.079
You know, if we use it's natural after all.
01:00:06.239 --> 01:00:07.679
Uh natural gas.
01:00:07.920 --> 01:00:08.079
Yeah.
01:00:08.239 --> 01:00:09.440
No, it that's wild, right?
01:00:09.519 --> 01:00:18.639
And of course, for people who aren't familiar, New South Wales, where Sydney is the biggest city, Victoria, where Melbourne is, you know, the two uh biggest cities in Australia.
01:00:18.800 --> 01:00:22.159
Uh no surprise that that's where so many of the data centers are going.
01:00:22.320 --> 01:00:41.280
Um, you know, there is a big push right now for moratoriums against data centers to stop the construction completely, at least until there are better rules and regulations and frameworks in place to account for the demand that we're seeing from these infrastructures, hopefully also to consider what they're actually powering and whether that makes sense.
01:00:41.519 --> 01:00:45.920
What do you think is kind of the approach that we should be fighting for here?
01:00:46.000 --> 01:00:48.400
You know, is there an alternative to what we're seeing?
01:00:48.559 --> 01:00:50.239
Should we just be shutting it all down?
01:00:50.400 --> 01:00:53.920
Are there ways that we can do this in, you know, a better fashion?
01:00:54.079 --> 01:00:55.360
What do you think?
01:00:55.920 --> 01:01:08.639
Yeah, I think we I think we as like, when I say we, I mean really broadly, like not just the climate movement, you know, but like anti-data center groups in general and you know, tech critical movements underestimated our own power a little bit.
01:01:08.800 --> 01:01:15.199
Uh maybe, you know, we've been a bit like sort of beat down over the past like half decade and we feel like we are powerless.
01:01:15.360 --> 01:01:21.679
But um the push for moratoria in particular has been remarkably successful in in many, many different parts of the world.
01:01:21.920 --> 01:01:32.400
Even if that push just ends up with like, you know, a sort of mild pause or like a new set of restrictions or regulations, it's still you know partly successful.
01:01:32.559 --> 01:01:38.239
And what that means, the most important lesson really is that like we need to actually start thinking about what we want.
01:01:38.400 --> 01:01:45.599
Um, because we were chasing after the car and we caught it, you know, it's like really uh shocking, but now it's like okay, we have to think about what we want.
01:01:45.679 --> 01:01:58.480
And and I think um that's why I spend so much time uh thinking about what it means to have an energy regulation of a data center, like you know, what what kind of law would you slap on a data center?
01:01:58.639 --> 01:02:13.360
Um, and and my contention, you know, uh I am I'm I'm always primarily in favor of moratorium, um, because I think they're the really the antidote to the impatience and the rush and the irrationality of the push behind them.
01:02:13.679 --> 01:02:30.159
But once you have that, um I think that basically having an extremely strict mandate for the construction of renewables and the basic guideline that I think that should be followed is like just look at the gas, you know.
01:02:30.400 --> 01:02:46.320
Nvidia is supporting this facility that literally is building an entire new gas-fired power station that is owned very, very clearly and instigated very, very clearly by that consortium of companies, and that is why we criticize it, right?
01:02:46.480 --> 01:02:48.719
Because we say, well, you're responsible for that gas plant.
01:02:48.800 --> 01:02:52.719
It's not it didn't just kind of spring up coincidentally at the same time, you did that.
01:02:52.960 --> 01:03:04.800
So so that level of ownership and direct causality that should be the absolute baseline for renewable energy mandates for companies, and they will they will absolutely scream about it.
01:03:05.039 --> 01:03:09.760
But the climate crisis is real, like the impacts are happening here and now.
01:03:09.920 --> 01:03:22.719
It should all be linked back to the urgency, the actual urgency, not the ridiculous hype narratives of like AI doom, but the actual urgency of dealing with um the the ever-ratcheting upwards uh impacts of climate change.
01:03:22.880 --> 01:03:24.800
So they'll they'll kind of moan about it, right?
01:03:24.880 --> 01:03:31.760
But it's like we are making you do this because we want to stop people from dying from the impacts of your data center.
01:03:31.840 --> 01:03:39.760
Uh, and it should be treated that way, in the same way that you would, I don't know, have like a seatbelt mandate in a set of legislation for when you introduce cars somewhere.
01:03:39.920 --> 01:03:42.400
Um, it should be treated pretty much the same way.
01:03:42.639 --> 01:03:45.519
This is a law to stop people from dying from your product.
01:03:46.000 --> 01:03:57.679
People in the energy space, I think, are not entirely ready for like for that, which is why I constantly bring up gas, because I'm like, they can do it when they want to, right?
01:03:57.760 --> 01:04:01.199
Like they can absolutely build whatever the hell they want when they want to.
01:04:01.360 --> 01:04:04.239
Um there it's under construction, like they're doing it.
01:04:04.480 --> 01:04:06.079
It's really not a problem for them.
01:04:06.159 --> 01:04:07.119
They can afford it.
01:04:07.280 --> 01:04:15.519
They just oppose it because they get online, you know, six months quicker if they build like some filthy gas turn turbine instead.
01:04:15.679 --> 01:04:36.719
The other thing that I that I really uh think is worth mentioning there is um it has to be paired with a re-establishment of like a central planning approach to climate change and energy, because you might end up there with a situation where like a bunch of companies are building renewables in a terrible, haphazard, community angering way, which we also don't want.
01:04:36.880 --> 01:04:44.800
There are different models for this, like you know, you can have the companies like put into a fund that that you know, and then the government looks after the power lines and the renewables and things like that.
01:04:44.960 --> 01:04:52.559
Um, but whatever way we do it, we need to have an overarching tracking view of how what is happening on power grids.
01:04:52.880 --> 01:04:58.960
There is a real complacency that has emerged in um climate and energy spaces.
01:04:59.199 --> 01:05:15.440
Sorry to self-criticize my own movement so much here, but I've had a week and I'm really, you know, um maybe it's the maybe it's just, you know, climate week, but uh we're not tracking in enough detail and with enough uh critical scrutiny what is happening on power grids.
01:05:15.519 --> 01:05:21.360
We kind of got a little bit complacent about renewable energy, and it's like, oh yeah, you know, it's just gonna grow infinitely and solve the problem.
01:05:21.440 --> 01:05:23.119
Um and it's not what's happening.
01:05:23.280 --> 01:05:32.000
Um, and the more attention we pay to that, the more we also kind of deal with this data center issue at the same time because, you know, they're occurring on power grids too.
01:05:32.159 --> 01:05:43.280
So that's why I'm constantly ranting about greenhouse gas emissions and climate change and you know, the measures of emissions from power grids, because it's not something a lot of people still talk about.
01:05:43.440 --> 01:05:46.719
Um, there's a feeling that it's kind of a solved problem, but it's but it's not.
01:05:46.960 --> 01:05:48.239
Yeah, no, definitely not.
01:05:48.400 --> 01:05:55.199
And I think it's important that you point out the piece around actually building this renewable energy in a way that is kind of in line with the community too, right?
01:05:55.360 --> 01:06:09.199
Because we have also seen the fossil fuel industry take advantage of, you know, renewable projects that are kind of forced on communities, you know, that are not done with proper community consultation and that can be used to turn people against uh, you know, renewable energy projects.
01:06:09.360 --> 01:06:18.639
And they can very much be done in ways that are in line with community, um, you know, that are even beneficial to community, again, if they're done properly, if they're done with the right intentions.
01:06:18.800 --> 01:06:38.320
Um, but Katana, I think you've given us a great picture of, you know, the environmental perspective on these data centers, aspects of it we don't always think about, including the chips, and why, you know, certainly we hear a lot about what's happening in the United States and parts of Europe, but it's important to pay attention to what's happening in other parts of the world because there are things that we can learn from that, like in Australia as well.
01:06:38.400 --> 01:06:40.559
Um thanks so much for taking the time.
01:06:41.039 --> 01:06:42.000
Thank you for having me.
01:06:42.159 --> 01:06:43.519
Always lovely to be back.
01:06:44.400 --> 01:06:48.159
Katan Joshi is a climate writer and data analyst based in Oslo.
01:06:48.400 --> 01:06:52.320
Tech Won's Save Us is made in partnership with the Nation magazine and is hosted by me, Paris Marks.
01:06:52.480 --> 01:06:53.760
Production is by Kyle Houston.
01:06:53.840 --> 01:06:58.480
Tech won't save us relies on the support of listeners like you to keep providing critical perspectives on the tech industry.
01:06:58.639 --> 01:07:03.679
You can join hundreds of other supporters by going to patreon.com slash tech won't save us and making a pledge of your own.
01:07:03.840 --> 01:07:06.400
Thanks for listening and make sure to come back next week.