September 14, 2023 – FS Insider speaks with award-winning investigative journalist Guillaume Pitron to discuss his latest book, The Dark Cloud: The Hidden Costs of The Digital World. In today's interview, Pitron provides a comprehensive examination of the materiality of the digital world we inhabit, including the surprising amount of energy consumed and used for the creation of semiconductors, the powering of massive data centers, as well as the training and utilization of artificial intelligence. As Pitron explains, the more advanced the technology, the more energy and material inputs that go into it. This is important when we consider the large amount of 'green washing' and publicity around the idea that going digital is equivalent to going green or, for example, when tech companies or others promote and discuss the topic of carbon neutrality. We highly encourage you to read Pitron's latest book and to hear what he has to say in today's eye-opening interview.
Timestamps:
00:00:00: Introduction
00:03:05: Impacts of Digital Information Transfer
00:04:03: Conflict and Challenges in Green Energy Transition
00:04:34: Contradictions in Digital World Advancements
00:06:20: Experience and Research Process
00:08:07: Energy Consumption in Data Centers
00:10:02: Evolving Raw Material Dependency in Technology
00:11:03: Importance of Graphite in Digital Devices
00:11:19: Working conditions and the reality of mining graphite
00:13:20: The unseen environmental impact of technology production
00:14:30: The environmental vs. digital industry
00:15:46: Understanding the impact of obsolescence in technology
00:16:22: The growing environmental impact of the digital world
00:17:00: Technical, software, and cultural obsolescence
00:18:28: The problem of consumption and digital pollution
00:19:51: Upgrading technology and the unseen cost
00:20:29: The energy consumption of creating a microchip
00:22:16: The underestimated power and cost of our phones
00:23:04: Business Model and Consumption Behavior
00:23:40: MIPS Method and Ecological Impact Assessment
00:24:18: MIPS Ratios and Material Legacy Behind Technologies
00:25:35: Connection to Data Centers and Artificial Intelligence
00:26:22: The Role of Data Centers in Daily Activities
00:28:24: Oversizing Industry and Ecological Costs
00:28:46: Energy Requirements of Microchips and Data Centers
00:29:17: Transitioning to a Greener Future with AI
00:29:44: AI and Deep Ecology
00:30:50: Risks of AI Exclusive Control Over Green Policies
00:31:56: Potential Problems with Green AIs and Energy Consumption
00:32:23: Choosing Between Infrastructure Development or Other Pursuits.
Transcript:
Cris Sheridan: Joining us on the show today is award winning investigative journalist Guillaume Pitron. He's the author of the must read book the Rare Metals War. And today we're going to discuss his latest book, which is titled The Dark Cloud the Hidden Costs of the Digital World. Guillaume, thank you for coming on and speaking with us again.
Guillaume Pitron: Thank you Chris, for inviting me again. Happy to be back.
Cris Sheridan: So Guillaume, I know that we touched upon this in our last interview when we were getting an update from you on what we were seeing just in terms of US. China tech wars and some of the various critical raw materials that are in the spotlight. You know, we ended that interview talking about how one of the big concerns when we think about the big explosion with the use of artificial intelligence and all these large language models that all these tech companies, the fang companies, are racing ahead to try to build out, is the amount of energy consumption that is required in order to train these models. And when you have millions of people now regularly using Chat, GPT, Google, Bard, and all these other ones for their use, there's a massive amount of energy consumption behind that that people don't realize and that that is going to be increasingly becoming a political issue that we're going to have, not just to mention an economic issue. And you had said, well, you were actually writing a book on that as we were speaking. So start off, if you wouldn't mind, tell us, what sort of led you from Rare Metals war into writing The Dark Cloud.
Guillaume Pitron: When I finished my first book, The Rare Metals War, I was mentioning a study saying that in the next 30 years, so until 2050, the humankind will consume more metals and minerals than ever since that has been consumed and extracted from the ground for the last 70,000 years. And the green world, because we need so much metals and minerals for green technology, is partly responsible for such a figure. And at the very same time, the words dematerialization virtualization, I'm going to put my paycheck in the cloud are popping up everywhere in our everyday vocabulary. And I found, like, there was kind of a paradox between the reality that we're going to consume even more resources in the future and these words claiming that we're going to dematerialize everything we do. And I went to see my editor and I said, why don't we do an investigation about this paradox of dematerialization? And this is how it started. And I wondered, how am I going to explain that everything that we do in the digital world has an impact? And I thought, well, what happens if I send an email or a like to someone who's sitting next to me in the Metro? In the bus, there is a 1 meter distance between me and this person. But will my data, my email, literally travel from my phone to this phone, which is 1 meter away. And actually, this is not what's going to happen, because the data, this email will travel all around the world, pass through 4G antennas, servers gathered in data centers, submarine cables, maybe satellites. And the real distance between two phones, even if they are 1 meter away from each other, is actually several thousands of kilometers, because this is actually the trail of an email. And the real distance is what I'm just mentioning right now. And this is such a paradox, which nobody's aware of. And so I took the pretext of sending an email to my neighbor as a way to investigate on what's very material in the physical world. And I traveled for one year for investigation around the world on the trail of an email.
Cris Sheridan: Yeah, one of the things that I remember was a key feature of your last book was discussing how just in order to see the Green transition happen, in order for it to be implemented successfully. I mean, we're going to need hundreds of new mines to extract all the necessary resources because renewable technologies actually do require vast amounts of copper, of course. Lithium for batteries, nickel, cobalt, all of these things. And this is the core conflict at the heart of the renewable energy transition, is the inability or unwillingness to extract all of these things that we need in order to successfully pull us off. And you detail that so well in your book. And in your newest book, you also show that there's a contradiction at the heart of the digital world than a lot of the people, organizations, industries, even these large tech companies that are saying that by going digital, we are going green. Right. Because there's this misperception around the cloud, like, as you discuss in the very beginning of your book in chapter one, the Digital World's Environmental Benefits, fiction versus Fact.
Guillaume Pitron: In fact, it's a wonderful technology, the virtual world, these digital technologies, because it makes us believe that we're going to be able to create more wealth, that we're going to get richer, that we're going to grow, we're going to grow our GDP. At the same time, everything that may have an impact in the real world, in the physical world, will not have any impact because we just do it online. The thing is, we only see these technologies through the best angle, which is what it brings to the planet. And we think about cost versus the benefits. And the cost is what materials are you going to need in order to make your phone work? An iPhone is probably featuring 60 different metals, which you need to mine somewhere. What energy are you going to produce in order to run the data centers, cool the data, 24/7. How are you going to lay 1.2 million cables all around the world in the oceans in order to keep ourselves connected? Twenty four, seven. And this is this that the telecommunications industry never talks about. And this is what I wanted to unveil.
Cris Sheridan: Yeah, it's so fascinating when you talk about some of the greenwashing or the green marketing that goes behind some of these projects and efforts. And many of us, we just kind of buy into this. But as you've done as an investigative journalist, visiting all these different countries data centers, looking at how semiconductor chips are manufactured and the amount of energy that goes into creating them, as you say, at one point up to 16,000 times the weight that goes into them versus what comes out. These are fascinating, fascinating realities that I think most of us just are completely unaware of. And I get the sense that as you were doing the investigation for this book and putting all this information together, that you were just as surprised as I was in reading it.
Guillaume Pitron: Oh, yeah, sure. First, this investigation took me to mines, because everything that is virtual stems from scar in the ground. We need considerable amounts, volumes of various metals, whether it's copper, lithium, cobalt, by the way, the very same metals which are necessary for making the green technologies possible. But these metals need to be extracted from somewhere. So that brought me back to some Chinese mines, among other mines in the world. And then I had to travel to Washington, DC. South of DC. There is a city whose name is Ashburn, which is nicknamed the cloud city. 70% of the world data produced every day is passing through every day, Ashburn, because this is a Silicon Valley of data centers in the United States and probably the world. And that is fascinating to investigate. That and also the electricity which is being needed for such data centers, which partly comes from coal, and such coal has been extracting the Appalachians. Fascinating story. Fascinating to realize that to cool the data, rather than using air conditioned systems, we use free cooling techniques because the air is fresh in Montreal or in Lapland, in the Swedish laplands. Facebook today, meta has literally fixed several data centers for consumers in the north of Lapland, north of Europe, close to the Arctic Circle, to have the data naturally cooling off. Thanks to the blizzard, especially in the wintertimes, everything we do online depends upon cables. And I've seen the deployment on cables on the Atlantic French coast. Everything we're talking about here is either invisibleized in the suburbs of our city, where nobody goes either underneath the ground. When we talk about cables, there are undersea cables. So everything is about infrastructures, not supra structures like a tower, but infra something which is below your feet, something which you don't feel, you don't sense it, you don't smell it, you don't taste it. And this is why it's hard to understand that this comes at a cost, because your senses are not being used. You don't use your senses to get a sense of this reality.
Cris Sheridan: Yeah, as you're saying it's, this invisible world behind us. And as you also detail in your book, the large tech companies are doing everything they can to make it that way, to keep it invisible and even using shell companies in many cases to build these very large data. Centers in ways where people don't realize that this massive building that's going up, and we're seeing increasingly hundreds and thousands of these data centers popping up all over the place in order to house and quantify and analyze all of this data. Of course, this is all driven by consumer behavior. If we think about the digital generation growing up right now, just how much time that they spend on their phones and on YouTube, on TikTok. So all of these data centers are being required because of where society is moving. And you touched upon something that I do want to highlight. You talk about how in your book, in the 1960s, a phone took about ten raw materials. In the 1990s, phone took about 19 different raw materials. Today's, cell phones take about 50. And of course, we've talked about a number of those different raw materials, critical raw materials of which are very hard to come by or require large amounts of earth to be strip mined in order to acquire these metals, unlike technologies we saw in the past. One of them you talk about is a very interesting story where you went to China. So tell us a little bit about your investigation and what you saw when it came to the use of graphite and the mining of graphite, which is obviously extremely important.
Guillaume Pitron: You wouldn't have a connected life, Chris, without graphite. Graphite is a mineral which is necessary for the battery of your phone. And 70% of the graphite on earth is being extracted in China and more precisely in the north of China, in a province. With them, is heliang Zhang? If you get there, you see the way people work. The working conditions are probably closer to what they used to be in our western countries back in the in the 50s than in 2023. And it's hard to film over there. It's hard to bring back some information over there because we're in China, because this is a country which doesn't welcome journalists. But what I could see, the reality of how we mine graphite for making our virtual lives possible, is quite shocking. And that was surprising. And we need to understand, Chris, that as you said, a phone is made of many metals, which are sometimes called rare because they're very much diluted in the earth's crust, but they have physical and chemical astounding properties. But if you want to make an iPhone, which maybe will weigh like, what, 150 grams actually, you need more than 1000 more minerals, more materials to extract from the ground and to use in the various refining process, including water, than the final weight of the product. Which means that you need 183 raw materials during all the life cycle of the product in order to make a 150 grams phone. So the real weight of your phone is more like closer to 200, be precise. And this kind of ratio, it's extremely high. And you can make such a ratio for every product in your life, a jean, a shoe, a pen. But you can make that for electric products, which are the most complicated things we've ever manufactured. And so it tells you that the more you want to go virtual by using such products, the more actually you need materials and the more the ratio will be high. So that brings me to the conclusion that, well, the more you want to go virtual, the more it is material, which is a complete paradox we didn't even think about. And we don't see this because we are not being talked about this because the iPhone is a beautiful object. How can you believe that such a beautiful object as an iPhone can be dirty? At the same time, we don't think about it this way, and we don't think about it also because, as you mentioned, the infrastructure is invisible. We don't see it. The data centers are very mundane infrastructures. There is no name on it, there is no brand on it. Nobody knows that it's Facebook or Amazon infrastructures. Sometimes the legal name in the books of the infrastructures of the data center is being changed so that you don't legally and officially link Meta and Facebook to this specific data center. So it's a way for you to see Meta and Facebook everywhere in the virtual world, but to never touch and grasp such a reality in the physical world. And so we end up not talking about it. And we should be aware of this anytime soon, because that's going to be a huge climate issue if we keep ignoring that.
Cris Sheridan: It reminds me of the quote, the greatest trick the devil ever pulled was convincing the world that he didn't exist. Like you're saying, there's this effort to promote this idea that the digital world and the companies that make the infrastructure for the digital world possible. I mean, they're digital, they're green, they're dematerialized, all the while ignoring this vast amount of infrastructure. And one of the things you just touched upon was the difference between at least if we think about this from an environmental standpoint, the difference of looking at the output through carbon emissions, which, as you explain, is incredibly reductionist, versus looking at the MIPS framework, which is all the inputs that go into that. And I really appreciated that part of your book. And when I was discussing this with others on a number of the key points you make, one of the pushbacks that I got was, yes, but the benefits likely exceed the costs when you put it all together. And one of the things that I think you make key on why that may not be true, why the cost may in fact exceed the benefits on digital is when you think about obsolescence. So can you tell us a little bit about why this is such an important part of understanding the dirtiness of the digital world?
Guillaume Pitron: It's very important to compare the cost and the benefits and to make an overall analysis. And thousands of experts have tried to do so. And at the very end of the process, honesty leads me to tell you that it's just impossible to know as a fact if the virtual world, the digital world, cost more to the planet than it benefits to the planet. It's just impossible to compare because it's just too complicated to calculate the cost and benefits and to compare. What I can just say is that I don't know. What I know, however, is that the environmental impact of the internet, of the digital world is growing by around 9% every year. And this is the fastest growing impact of any industry comparing to any other industry that we are witnessing right now. And this is becoming an issue. It's becoming an issue, especially when, and I get back to your point, we take into consideration the plant obsolescence. Basically, your phone is in a perfect world of the phone manufacturer. You won't keep your phone too long. You'll have to change it either because you have to because the battery is down. Usually the battery is the first thing to go down, and you can't change the battery because the battery is glued to the rest of the phone. So it's like a technical obsolescence where you can't change your phone. You have to change everything because there's just a part of it which doesn't work, while the rest perfectly works. This is an obsolescence that makes you consume more phones. Then you have like you call the software obsolescence. And you and I have already had this painful experience of not being able to download the latest application because it was said your phone is too old to download the latest iOS, which is not compatible with the latest app. So change your phone, get a new one if you want to be able to download the latest app. This is software obsolescence, and you have the last obsolescence, which is called cultural obsolescence, which means basically that you have your iPhone 14, which about last year. But as you know, the iPhone 15 is about to get to the market, and it has a better battery, a better camera. It's just more beautiful. And even if your past phone perfectly works, you just want to change it because you want to get the latest one. And that makes you consume even more of this resort of these interfaces tablets, phones, computers. Today, there are about 34 billion of these devices all around the world, and they account for 70%, 70% of the world digital pollution. But think about what we could do if we could keep our phones just twice as long repair them, resist to this cultural obsolescence that would change things, but we're not ready for that.
Cris Sheridan: Yeah, and companies are making it they're intentionally trying to make it so that you cannot replace certain aspects of the phone, the battery, for example. Like you said, they're glued to certain portions. Or if you were to even go in and try to replace certain items, then you either ruin the warranty or you could just end up breaking the phone completely. So obviously, companies are doing everything that they can. And on top of the planned obsolescence, which is a thing, you also touch upon the cultural obsolescence trading in your car every two years. Or like you said earlier, I mean, upgrading to the newest iPhone, not because your old iPhone doesn't work, it just perhaps doesn't have the latest technology. Right. Then there's software obsolescence. So obsolescence is a huge factor, again, for why we have to understand some of the trade offs that are taking place and why that tips us a little bit closer to the idea that the cost may exceed the benefits of going digital and perhaps being more eco friendly. Because, for example, if in the past, if you had a 1960 phone with ten raw materials compared to a cell phone today of 50 raw materials and that 1960s phone you kept for ten years versus today's cell phone, being traded out every two to three years, especially if you're a kid and you're getting one at 13 years of age, and you're probably going to break it in a year. Well, these are all the considerations that really factor into the analysis which you covered so well in your book and I really appreciate that. We do need to talk about before we go, we got to talk about microchips, because obviously, scaling from microchips to data centers is a huge part of understanding the core of your book here. Can you tell us about just the amount of energy consumption that's required to create a microchip?
Guillaume Pitron: You need 16,000 times more materials for producing a chip than the final weight of the chip itself. So if a chip is 2 grams, in fact, you may need up to 32 materials in order to produce a chip, which is the highest MIPS material input per service unit ever known of human history. Which tells us, once again, this very reality is that the more you want to go virtual, the more you need to get materials, and the more you go material that tells you how much virtualization is a myth, is a hoax, and is dangerous. Part of this figure might be explained by the fact that you need energy in order to print the circuits on the microprocessor. And this energy may be produced out of coal or out of oil or out of metals for a wind farm, but that comes in in the final weight of the product, because you take into account in this calculation direct indirect products that eventually make such a chip possible. Now think about what you do with this chip first. A chip today is much more powerful. A single chip in your phone is much more powerful than every digital devices that have helped send men on the moon back in the it's much more powerful just in your pocket. Think about what we could do in the 70s with such power and what do you do today with such a phone? What do you do? And the big question here is do you need such a power in your phone? Do you need to have such a power to take selfies and send videos to someone which you will forget the next day? And I think there is a big question here which is never being asked, never considered, because no one wants you to consider such a question, which is what is the true cost benefit analysis that you make regarding how much resources, how much materials, how much energy you need for the real use you make out of it? What is the environmental cost of it? And how you compare that to the social, the political, the individual, the psychological benefits of having such a power in your pocket. And this cost benefit analysis has been the key for men to survive for whole history of humanity. And we don't think about it anymore because we think it has no cost, because it's free. Because the basic business model of the Fong is to have the devices free in exchange for your data. So you don't realize the cost of it. And we end up with devices which comes at a cost which is unnecessary comparing to the real use you make out of it. That brings a question of your consumption behavior, of the right balance to have in this world which is in reality constrained. And this is maybe time for such.
Cris Sheridan: A debate and touching upon something that you had just mentioned. The MIPS method is basically looking at the number of inputs or the factors that go into producing an object. So you can look at the weight, the collective weight of all the energy and the logistics and the mining and so on that went into producing an object versus what comes out of it on the end, right? So it's not looking at carbon emissions, which is a more reductionist. The MIPS is a much more holistic framework for assessing the ecological impact. And obviously this touches upon a number of other areas as well when it comes to geopolitics, as well as what you detailed in the Rare Metals war. But one thing that you write here is that the MIPS ratio of a television, so again, this is the amount of inputs relative to the weight of what you get out is 200 to one. So 200 times the weight of what went into it to produce one television, it can be as high as 1000 to one. That of a smartphone is 1200 to one. And then you say the price for the highest MIPS goes to the microchip and that is 16,000 to one. So that's extremely high given what you get out of it. And you talk about you visited a microchip processing facility, just the amount of water that is required, which is an extreme amount of water, and that is becoming an issue for a number of different places, not the least of which is Taiwan, one of the largest. And I want to quote here, you say that these dematerialized, in quotes, dematerialized technologies, not only consume materials, but are quite simply becoming one of the most massive enterprises in materialization in history. And then you go on to say that the uncomfortable truth is that the more wearable, discrete and light the technology, the greater the material legacy behind it. So those are some really key points. So scaling it up as we close here, let's end with, again the data center and how this fits into artificial intelligence. But the microchip 16,000 to one MIPS ratio when you consider all the energy, logistics, water, raw materials that go in to produce such a small thing and then you have to multiply that by billions for all of the data centers. Just one data center in particular consuming enough electricity to power a small town, if not a small city of 10,000 to 10,0000 people. You add all this up and you're looking at a very massive, energy intensive infrastructure here. So tell us about what you learned about data centers in particular.
Guillaume Pitron: What you need to understand. Chris and what we all need to understand is that each of us individually use every day 100 data centers. When you send an email, there is a data center. When you swipe on a dating website, there is a data center behind which sends a swipe to the data center, and the person you swipe on it will receive your notification. When you order a Uber or pizza, there's a data center. But in fact, there is much more than one data center for everything that you do. But probably 23456 data centers. Your Gmail account, if you have a Gmail account, actually is scattered the same data replicated in six various data centers. And you may ask, why is that? Well, because basically, you need to have the shortest latency possible, which is the shortest distance between you and the closest data center possible, wherever you are in the world, in order for your page to connect, to download as fast as possible, because you can't wait more than 1 second or 2 seconds in order to have the page downloading. So you replicate the data center to always have a data center which is as close as possible from where you are in the world. And the second thing is, if a data center goes down because there is an electricity crash, for example, you need to make sure that there is another data center running at the very same time which will make sure that you can surf on the web 24/7. So you need to make the infrastructure available all the time. The sun never has to set on Internet for you to be always online and to be as quick as possible. So this is my behavior as an internet user because I can't wait anymore because I can't accept to have Gmail telling me, sorry, we are running out of power. Come back tomorrow to check your email. That's just not acceptable anymore. The industry has to oversize everything and to replicate everything so that you can be online all the time. And this comes at a cost. And what comes at a cost from a pure ecological standpoint is to, as a consumer, want to have everything everywhere, all the time, as fast as speed possible. And that comes at a cost.
Cris Sheridan: So, as you explained, microchips require vast amounts of energy as well as data centers, which are increasingly putting competition on towns and cities for the amount of electricity that they require. And as you discussed there's, I believe over 3 million data centers globally. This number is climbing very quickly, especially as we consider the amount of supercomputing that's necessary to train these large language models, which are also growing very quickly and becoming part of everyday society. And the applications. Towards the end of your book, you get into a very fascinating aspect, and this is a little bit more philosophical, but of course many people are saying that AI will be the solution to essentially helping us transition to a greener future. And you have a question here about whether or not we would be happy if artificial intelligence were to embrace the philosophy of deep ecology. Do you mind explaining that little portion of your book which I found really fascinating?
Guillaume Pitron: Yeah, you're raising a fascinating question today. Some AI are said to predict the weather, predict the CO2 emissions in a Chinese city, according to the weather, and to precisely tell you what you exactly need to do three days in advance as a policymaker in order to prevent the excess of CO2 in the air. So basically you have to do nothing. Just push the button and listen to the AI. And this device, which have been promoted by American companies, already do exist. Now push this logic further. We, the human race, species, are just unable to fight climate change because we don't want it and because it's too complex. What if we give a general AI the power to decide of our green policy worldwide for the next ten or 20 years, because such a device would be much more clever than us and will be able to take decisions which we cannot take on ourselves. And what will be the philosophical risks? And these risks have been considered by researchers at Oxford University in Great Britain. For example, what will be the risk of letting. General AI take in charge our Green policy for the next 20 years. And what if this AI may actually decide to suppress the human species as it would be maybe the best way to protect the planet. So we may have developed in that situation an Earth friendly AI. But that wouldn't be a human friendly AI. And how do we make sure that such a situation doesn't come reality? It's science fiction today, but philosophers, including the United States are questioning that and just want to make sure that saving the planet shouldn't necessarily be a reason for doing anything irresponsibly when it comes to developing AIS and especially Green AIS. This is fascinating, what I've discovered when investigating this.
Cris Sheridan: Yeah. And I really enjoyed that part of your book. And you talk about a green leviathan against humanity. And of course this all ties into the idea that microchips data centers and these large language models AI collectively require massive amounts of energy and they actually are now competing directly with cities for power consumption and increasingly the very critical raw materials that are being used in a lot of these devices. And the goals that we have as a society either to further develop this infrastructure and this always days on internet, the cloud, or if, whether or not there are other things that we should be pursuing and plowing more of our money into and our time and resources. So, such an excellent book. I highly want to recommend it.
