Gauthier Roussilhe 'The hidden costs of data centers'

This transcript is auto-generated by AI and may contain small errors.

00:00:00

S1: Got you. Rosalie, a specialist in the environmental footprint of the digital sector for five years and is certainly among the most knowledgeable people I have been lucky to chat with about these issues. He's currently doing a PhD at RMIT, that's the Royal Melbourne Institute of Technology to explore which digital infrastructures and services are compatible with a world stabilised at plus two degrees Celsius. I started our chat by asking Gautier about the electricity impact of data centres.

00:00:38

S2: Well, historically, there has been two assessments of how much datacentres consumed worldwide in terms of electricity. There have been the IEA estimates, so running 200 terawatts hour, which is based on the work from Shelby and the hall in the US that are that are center sector is the analysis was made in 2016 and from this first analysis was then derived the IEA analysis, but also the Massenet and Kumi analyses that have been widely used so far. Also assessments comes from the bulldozer models, some institutes in Germany that did somehow the same analysis at the EU scale and also the worldwide scale which very different numbers and scenario for world wide data center consumption was around for 400 terawatt hour. So the best we know so far from a worldwide perspective is that the range of electricity consumption for datacenters is around 204 countries. To be fair, I don't believe knowing how the datasets are being set up in India from the US side, I don't really think that is the most realistic one because it assume that large parts of the worldwide data centers would go hyperscale and they extrapolated that's market penetration of hyperscale worldwide to get to number of to 200 with many other factors. But at the end of the main hypothesis behind the work, when the German estimate is considering a less penetration rate of hyperscale in Europe and in other markets, and and if we if we look at data coming from other countries, but that cannot be verified, sadly, because that is not open. And that is a the methods used are not open. But for example, the Ministry of Telecommunication of Technology and Information Communication in China is estimating that under the center energy consumption around 200 terawatts per hour, which doesn't make sense according to the estimates that we have. But since we cannot see what they including in that what they are going datacentres, we actually don't know what's happening in China and the thing is that. From from the IAEA estimates and based on shabby and let's put it the extrapolate from America that said to the world and the from the budget websites extrapolated mostly German data on European data to the world. So it's actually very hard to have a precise picture. And that's why we always need to get a more territorial perspective, to actually look specifically in countries to see where what does it actually means locally, and to set up policies that are that are effective towards a specific situation like in item.

00:04:06

S1: So in Ireland saying, you know, 14% of Irish electricity, it's an enormous quantity in a very short period of time like in in the history of datacentres. And it's you know, people are estimating it'll grow to 30% if if the growth of data centers actually continues. So in in country environments or certain country environments, it's it's a significant element of the electrical. Next.

00:04:37

S2: The issue in Ireland is capacity. We need if we want to grow so that that centre sector that much in Ireland, then we need to increase capacity. The fact is, if I understand where the full scope of the power provider, our grid doesn't have the enough possibilities to. To increase its capacity according to the demand of that the census and also to the other sectors demand. And this issue is not new at all the same up in Amsterdam, in Frankfurt, in Singapore. So these kind of planning issues are always up and is because we have a sector with a large demand increasing very fast. And that is unsettling for many poor operators. We have this also same issue in France, in the north of Paris, where basically you have to make a choice. That's you either follow your energy transition roadmap, which does not include that you need to reduce your energy consumption a lot or you follow economic developments pathway which will always favour data centers. And I don't know how it is in Ireland, but I know like in France between now. So I mean, let's say 2020 to 2050, we have to we have to reduce refinery energy consumption by 40% in so so in 40 years in 30 years, we need to reduce by 40% to all energy consumption, which means going from 1600 zero, what the hour to 900 hour by 2050. And that's actually the framing that we need to have in mind when we are analysing such sectors. What's his energy transition roadmap and does it fit into it? That's the main issue because of the added moments you can have that the sentence operator an item that will absorb all of the renewable capacity being put on the grid and not allowing the sector to get it at the same time increasing the demand for electricity. So we end up with that. The centre's operators having a very nice CSR reports with lowering the lowering the carbon intensity of the electricity mix, but not allowing other actors to get it. So again, if you look if you look at it from a Toyota perspective, it's a zero sum game.

00:07:13

S1: As you said, the renewables that go to the data centre account go to older. You know, there's a was there's a limited capacity for those at the at the moment in the process. So, you know, but the bigger picture or maybe not the bigger, but the electricity, it's part of of the overall impact of a data centre. But the story that often isn't told that got much is its physical infrastructure. And, you know, could you give us a bit of a sense of all the other elements of the data centre that that we need to focus on from the environmental point of view?

00:07:57

S2: So if you take it from a lifecycle assessment perspective, the first thing is the construction of the building itself. So sometimes operators will take will take on the old buildings and refurbish, then renovate them according to the specificities that there is that the centres. Then you need to include all energy equipment and cooling equipment. So. Basically. AC if it's if they're using dots of water, what what's a cooling system? And also the fact that you will have you will be plugged to the energy read and you will need translators and so on to gets to get the rights those are rights. Electricity in but also of the energy system that is out here in case of power blackouts. So which means first batteries. Lithium depends on the technology, but it can be lithium batteries that will take on for work that we take on in case of blackouts and keep bringing electricity to the data center for a few minutes for the engine engine to kick in and to provide a stable power line, which means so in number in each of the sensors, there is always a turbine like basically this like a like a motor for for fold. But we for fuel we for fuel reservoirs at least to maintain them that are centos depending depending on the operator from 2 to 4 days, maybe even bigger. So that's a things you start with and then you have building with electricity coming in, you have good cooling coming in and then you need to put your hardware in. And but to be fair, from an economic point of view, the most important part of the building is hardware. It's not the building or the energy systems, because you can build your hardware anywhere. That's where you are, your value. The building itself is secondary. Obviously, it's not easy to to move all your hardware from one center to one another, but it's possible. And then this is all the work on from with a material footprint now. So those can be quite carbon heavy, but also it depends of the of how much time you will keep them. Um, datacentres on average tend to like from when we are doing like psychoanalysis, we had thinking the hypothesis is that the server will last at least for 4 to 5 years in its what manufacturers are providing as has as a reference point, then it depends in reality in the big hyperscale, especially from Google, Amazon and so on, we have like whisperers are in for that. So refresh rate for the servers in such facilities is maybe around 18 to 20 months, which means it will be discarded right after. It just means that they will enter a second cycle as refurbished. And that's when you as a is such a big, big market for refurbished hardware adware because there is being dropped by by Google, Amazon and so on. And yet if the server is normally for what, all the time being being around 24 hours, seven, seven day a week, that most of these footprints will come for electricity consumption normally. So standard ratio is that the material footprints represent 20%. I mean, the manufacturing footprint represent 20% of the overall footprints. And so use Facebook for consumption groups down to 80%. But it's based on the scenario your hardware last for five years and is not refurbished in between and uh, and then there is all end of life thing that I don't really, I don't really know because I only know about free for refurbished markets, for server, for datacentre on the web. But I don't know how that would be interesting afterwards. One selfishly, Doug and I don't actually know how long a server can last because always it becomes operation you are putting in.

00:12:41

S1: It seems that the manufacturer server can be in the region of a ton of CO2, you know, somewhere somewhere in that region. A short life for these products is in. Is very negative from a manufacturing point of view. You know, if you're constantly going through that cycle and some of them do go into refurbishment, but some of them are some of the components stay. They get physically trashed, don't they, from a security point of view?

00:13:13

S2: Well, for hydrides, yes. And then for ICI or for integrity, security, it depends. And at the end, what we get from that, from a recycling point of view, it's little, little, little minerals. It's a mix of aluminium, copper, and something called in French little. But they don't know how to say this. So we are losing most of the materials. I mean, from most of the time when we are dealing with waste going from the steel sector, we are we are recovering very few metals compared to the sort of diversity of metals that we have.

00:13:57

S1: Yet I saw one study that said with all of the e-waste that is is recycled, often only about 30%, 30 to 40% of useful materials. I actually returned from the actual recycling process.

00:14:16

S2: Yeah. Yeah. It's because I mean, if you just take a mouthful of things, there's like a like 50, up to 55 metals in the in the smelter with very different volumes going from mg to two grams and most of the around metals. Also, small metals will never be recovered because sourcing so small quantity or being using an alloy or just cannot be taken back. So we are losing most of all rare and small metals when we are recycling ICG products.

00:14:51

S1: So we've got the hardware, we've got the server hardware, the computer hardware, but we also have an electro mechanical hardware that is often has a much shorter lifecycle than it would have in a in a factory or a you know, the other is, you know, I read that about a 15 to 20 year life cycle for the electro mechanical hardware, whereas in a typical factory that would be expected to be used for 40 years.

00:15:22

S2: Oh, you mean like generators, batteries.

00:15:26

S1: Generators, air conditioning. But yeah.

00:15:29

S2: I don't I don't know that much on that. But the thing is, we know it lasts less in datacentres, but it's also a small part. So I mean, until today and to my knowledge, it was a small part of the footprint of the footprint because it's the one time big equipment with very low, um, I mean we've relatively long life, but I don't know that much because in LCA we don't include the, of the type of methodology, um, the construction of the building and putting inside the electromechanical equipments is very, very, very rarely accounted for. So actually I don't have a clear picture of that because we don't have data on that.

00:16:26

S1: You know, even the buildings are only supposed to last 20 or 40 years. That is a kind of part of this whole digital culture. That's that's nothing is designed to last.

00:16:37

S2: The same lifecycle assessment we take. So I put this is that's a building we lost 20 years basically you get your return after if I remember well 5 to 7 years of operation. So after 37 years you can consider that you can you can move out if necessary. And you can depends on many factors like what would you move your datacentre oxygen can be like. That's very could deal with electricity somewhere else. It can be also environmental issues. Like for example, I've never really understood why big hyperscalers were installed in a high water stress region like the west part of the US. And I don't see how this kind of operation can last over more than 20 years because a lot of stress in Colorado is already. Increasing quite fast, and yet we have more and more datacentres being constructed in the region. The same goes in Texas and goes in Utah, so I don't really understand it. What does the mid to long term perspective for datacentre operators and the effect is be on the whole is we keep building so much more datacentres with colocation, hyperscalers or even more like with edge. And I still really don't know how we are building that much regarding the capacity that we already have and the demand that is currently there. So my guess is that we keep making new data centres because it's a safe financial asset to invest in and also because it's based on the perspective that there be always more services being put in, more computing power being needed and more transfer and storage capacity needed. But yet it's just based on the perspective. It's not something that is taking into account. So the fact that we might also need to. To write correction or to to stop the development of not to submit to at least to slowdown the development of the two centers worldwide.

00:19:01

S1: There's a couple of things you brought up there. The water crisis that global I saw this morning, an announcement by the US government that the for the first time ever it's treating water as a national security issue, which is a historical moment to some extent. But as you've indicated there, the western and southwestern part of the US under historical drought conditions as are significant other areas in the world and the UN predicts, you know, very significant droughts by 2030 onwards and crop lands up to 80% of cropland has been stressed. And and yet, you know, I saw this study I think by op time last year, which said over 60% of data centers don't even bother to measure how much water they are consuming because they don't even think it's a business issue. They don't even it's not even it's not even worth measuring. And yet we know that that data centers can be can be quite water intensive.

00:20:13

S2: Yeah. I mean, it's a very specific case for the West and both of us have been trying to track down so what the footprints or this or what's a dimension of the distance operators in that part of the US because of the drills that they were facing in July, August 20, 2021. And first, it's just very, very, very hard to get to the data was asking for asking for water. Like most companies, datacentre operators like Facebook or Google. Amazon said don't use the full name. I mean, they still do the official putting them to such events and they use screen companies or so all the that are in charge of data center development. So first you need to find the name of the company, of the screen company doing this. Then you need to find the dates of the actual meeting of the municipal committee. The wells are deciding how much water like what's going on, the water demand of the operator. I did that once for one datacenter in Utah in the form of Facebook, Eagle Mountain Data Center. It took me 4 hours to just find one that points which is which is as a starting of their operation. Facebook will ask for 1000 kilometers of water and can go up ten. If I remember well, a 10,000 square meter cubic meters of water was a finished development. But on average, the Google Hyperscale at the center of. We love the what's our demands of? 15,000 kilometres of water bottles. This water is recycled, recycled, reused and so on. And it's not all of the freshwater being used to need fresh water to put on their datacentre. But the bigger, bigger picture here is that most of the water footprint of a datacentre is in the US, especially in this region is indirect. It goes from electricity generation. And if I remember well it was 80% of the of the water footprint of datacentres is indirect. Because producing power has also water intensity. But we need to work on for getting from dams or for students to students of mine. And it's also a price pressure because of the state and local world demands. And the more there is conflicting views about this power and also on the water as it being used for disposal and also locally on the water, is it being diverted to that centre and that will not go to other activities. So we can at some point it will not be surprising to have conflict of views in California, for example, over water, over water demand between farmers and and the datacentre industry. And most of the time, it is not the photos that are winning the kind of conflicts.

00:23:30

S1: This kind of leads to. You know, my reading about data centres. You know, when I read from researchers or people like yourself, it's almost a universal response that the data centre industry is extraordinarily secretive.

00:23:50

S2: Oh yeah.

00:23:51

S1: And, and, you know, I read that water is wet from a Google perspective, is considered a proprietary trade secret. And they they won't even allow public officials tell the local community how much water is is used in their use using. So why is it that you know, it's a kind of the irony, isn't it, of data centres? They are the places that store all the information, much of which about our personal lives and what we are doing on a day to day basis. So data centres and the Googles and the Facebooks have this extraordinary understanding of our lives, and yet we know hardly anything about them. And they, they, they go to extraordinarily efforts to be secretive. What are they hiding or.

00:24:51

S2: It's it's, um, I mean, it's a complicated question, but like many people that work on data centres, footprints. We all always face the sort of secretive nature of the business. It's very hard to have data and to actually that does it is good enough to be exploited in in research. So most of the time you work with whispers or things that you hear when you go to professional events of this industry. But yeah, it's hard to get. When it comes to Google, what's a footprint? Yeah. No, we are. Only what's left is the. What a footprint of the all cowgirl company. So it's an aggregated version of the water from Twitter, which is quite useless. I mean, if you're looking from a total perspective and trying to set up policies at the scale of your state or county aggregated what a footprint of Google is useless for doing the national or that is this can be can be used and also it will become I mean for me it will be it's starting it will be plug books to even more. I mean, in some in some facilities, I mean, some some states. Google is also building its own pumping station. So it will be even harder to know what's been extracted. And and even on the power on the poor side, Amazon is also building its own.

00:26:27

S1: Power station or.

00:26:29

S2: Power station substations.

00:26:30

S1: Substations.

00:26:31

S2: Yeah. So so you can also see like a virtualization of, of the business because they'll also starting to develop with commodities such as water power. And the more that it gives us in the think tank of the US operation on that and the less we're going to have access to it and that's a lot of transparency, that's a material source, local, local conflict. I think the system to hide it obviously can be said it's a states, it's a it's like a trade secret and so on. But it's also because I mean, my hypothesis is that the centers are facing more and more local opposition worldwide by local communities because they start to understand that's I think the sense of their buy comes with a price and most of the positive impacts. Are not primarily primarily seen by local communities. When you think of hyperscale settings and searching rural communities in Utah, Colorado and so on, most of the direct employment for local communities will be security, training and so on. In Israel, 12 local people are training to become I.T. engineers. It happens, but it's it's not for locals. And and they come and also this kind of operation come with a lot of fiscal or financial incentives. So even data center dynamics like specialized, specialized websites, and that was making a joke of the Facebook eagle that a center was claiming that they were contributing to millions to local communities was it gets a tax cut of $150 million. So this also I could just proportion of that which also leads to local conflicts that then will that that will result from that will result from an increased secret secrecy approaches operations.

00:28:49

S1: Now, as you've indicated, the data center does very little for the community. It's in in from employment or otherwise. It's not like a, you know, another traditional industry that becomes part of, you know, the community if it's a supermarket or, you know, a dentist or, you know, does some. Most of the data is not being used by the local community, so to speak. It's going all over the world or whatever. So when a data center comes into a community, it takes far more than it gives.

00:29:25

S2: And it's not a labor intensive industry to start with. I mean, and also parts of it are becoming more and more automated. I a remember there was a big controversy in the north of Paris in 2013 that was touted by a sociologist called came on market. Well the city of Paris kept related that the employment rates for data center in the in the in the area was one full time employee per 10,000 square meters was the average in the in the in the city was around 50 for the employees then 10,000 square meters. So it's so way less job intensive.

00:30:20

S1: Day it is growing at an extraordinary pace. But according to your understanding, data centers are growing at an even more extraordinary pace. Are we are we essentially potentially looking at a financial bubble type A situation with data centers over the next 5 to 10 years?

00:30:39

S2: Well, to be fair, I have no idea on that because it's I don't think that much. I mean, I'm only following from a certain distance the economics and the financial news around data centers. There is a good newsletter to do that. So one from Christian Kush that's that I receive weekly and you can see like a lot of financial operation happening in Seattle. It's quite insane. And, and yeah, for me from, from understanding I don't understand. How we are, why we are building so much water centers and also because. We are building some data centers where there is already data centers. So it's not an issue of support. So some data centers could be built for an issue of geographical coverage, which is understanding. But when you look at the map of data centers, they are all concentrated in the same location and which is around big cities most of the time. And these are all at the same place. So it means also that the demand is already already quite covered. So why are we building more is a question. And I think of you see some forecasts about new demands of computing, storage and so on and transfer is something. But I know I think also it's also from a way to consider the financial aspect of it as a safe assets to invest that and datacenters operate as a safe assets to invest in. So with steady growth, most of them at least, and everything is working quite well for like if you think of Equinix, if you think of stability in both recently the huge operations and they're still expanding very, very fast and in also new areas but also consolidating in many other places was they already are. And also at the end that you can see that there is only there will be only few companies left. So there is a concentration of power for datacentre operators like colocation, like Equinix, etc. and so that can be the same as the ones you can see for full service industries such as Facebook and so on.

00:33:08

S1: Yeah, there's this overall concentration that that is is going on and in the tech industrial has been going on for for the last 20 years just, you know, so we are at a very basic level. Data drives, data centers, our processing of data day drives, data centers, or you would think should drive the growth and in data center the growth of data. But you were talking about, you know, some of the articles I read, you know, that you've written about the the role of data or measuring data or that the consumption of data and that you think it's not always directly correlated to. More data is necessarily more energy consumption, so to speak. Could you could you just explain that a little bit, tease that out a bit?

00:34:14

S2: Historically, we use data traffic as a proxy to and Q4 to forecast the so footprints of the sector. But we did that because we didn't have a better proxy. And in a very secretive sector that's it's not because you're increasing traffic. That's why we increased footprints. If you just think of direct consumption, the only places where data transfer is increasing consumption and in a no nonlinear way, it's access networks. When you are basically transferring data for GMT station, but it's not like maybe 20% of the energy consumption of the station. When you think of data centers, it's not because more data is transferred directly, so electric consumption would increase for fixed access network. It's the same whatever you you put a lot of that downloads is not influencing that much consumption so it's not easy to use data as a proxy to estimate the cloud footprint. You can use it as like when you look backward once you know like how much to sell because your network and how much that interest compute, then you can do that. But you look at it from an economic point of view, it's because you decide that one gigabyte is the economic unit that you're looking for. Less like a typical operators can do that. We can do forecasts out of those because if you think of it's that that transfer will only affect renewal or renewal of public consumption hardware such as smartphone or adware or smartphone or computer when there is a change of generation of smartphones, and that's pretty much it. When you look at server and data centers, it doesn't affect directly. If you look at networks, it only affect directly the artists directly in of the Remote Access Network, but it affects indirectly. Then you need to distinguish what is data transfer from a physical point of view and the whole affects all infrastructure. So you need to distinguish that transfer from as a physical phenomenon that affects an infrastructure from that transfer as all data traffic as a narrative to justify a change of mobile generation. You can say 4G networks are saturated. We need to upgrade to 5G. Well, that's that's that's the narrative that's based on data that says we need to increase capacity.

00:37:21

S1: We were saying there earlier about how in France and Ireland, you know, we have to reduce energy consumption over the next you know, it's not about find. It's not an energy production problem. We have written energy consumption, one, if we're going to have a liveable planet. But but let's say if we you know, on one side, we've got this explosion, what Internet of Things or whatever. But if we if we broadly speaking, said we could reduce the amount of data by 40% or, you know, like the electricity consumption, you know, at a macro level, that would have an impact, wouldn't it? It would. You know, basically, if the the the data centers are growing, we're in the hope that we're going to have massive Internet of Things data in the process. So whereas at a gigabyte level, it's hard, you know, to track, but you know, at a macro level more data in does that at the zettabytes level the explosion in zettabytes of data. If we could reduce that by 40%, that would be a good thing.

00:38:30

S2: Well, if we say that's so, all volume of the traffic will be reduced in ten years. Well, it will directly it will obviously affect the way we are building capacities. We'll build less datacenters because we assume that the focus will be negative in the future. And so we maybe we also reconsider did things a new generation of networks. But I have my doubts on that because like we've been pushing for. You sing in front sing like 4G saturated. So we need to we need to go to the next generation. But if it was not for that, they will have funded this narrative and luxuries or I mean, the way it was. The folks we are pushing, 5G, is mostly to to deal with two massive, massive machine connections. So I think it would just be a change of narrative to keep developing a sector that is profitable. So and so at the end, the question is, can the sector be profitable? We keep going profitable. We have less data to handle. And and actually quite an interesting question because most of the business model of, let's say, Google or Facebook is based on data collection providing and getting better data and so on. And so it's and even the way we are training models is based on massive of massive use of, of sorry use of massive datasets. So the way we are developing statistics is based on the abundance of data. So could it be profitable if were based on that's immensity of data production is a good question and to be sure, it would be correlated with energy consumption. If you if we reduce the capacity of our infrastructure or even if we standardize it because if even if we stabilize the sector as it is today and build way less at the centers and keep it in networks as they are for a while, with the efficiency gains that we have, we could actually be able to to reduce energy consumption. But to be fair, I think that's not a big worry of the sector, because I think most research on the topic agree that energy consumption of that aspect of the sector, especially with activity, will increase largely in the coming years. And so and they will know was that some of the emission by integrating way more renewable energy and also by buying compensation and using mechanisms such as power purchase agreements to guarantees of renewable energy. But the fact is, we any way we need to reduce our energy production because we also need to reduce premature footprints and renewable energy of more a much more intense when it comes to the metallic intensity. So if we will reduce single energy consumption, all for another magic assumption is also reducing the limited intensity.

00:42:02

S1: A lot of my work over the last 20 or 30 years has been working in organizations, big, big organizations, and helping them better manage our websites, our Internet of Things at that. And in a pattern I kept noticing again and again, this is from 96, 97 onwards, is that basically 98, 90% of the data in an organization was crap, was really, really low level. The Internet's worked better if we deleted 90% of the pages where wherever I went, whatever organization I worked with, there was this massive quantities of really totally useless data. And and the more I've researched this area of, you know, studies that say that only about 5% of data is actively managed on a process. So anywhere, you know, from 80 to 90 to 95% of data after a couple of months, even most of it when it's collected, has no reason to be collected in the process. So maybe that's partly why data centers are extremely secretive as well, because they're actually really data dumps and they're they're pretending to be these wonderful places which store tremendously useful stuff. But really they are they are they are dumping grounds for the 1.4 trillion photos that we take a year, the 99% that we will never look at, look at again, and that we really do need a review of data, because if we have less data, as you say, I mean, we don't need 5G. 5G is a marketing, you know, other than, you know, factories and medical institutions, the vast majority of people we can get perfectly by and 4G in in the process. So we are actually getting new infrastructure not because we need it as a society, but to meet superficial wants of downloading etcr videos that we can't see the difference between an K video and a2k video in the process. So we actually have got into a cycle of data for data sake or data for the sake of making money for the data centers, even though the vast majority of that data is extraordinarily poor quality.

00:44:23

S2: Mm hmm. I mean, from my perspective. The fact is, when we are talking when we are talking about the descent of development and so on, or even that our production is that we are missing the framework in which we should evaluate Cisco things is the run up for energy, transition, transitions, all the problems that we should actually follow if we want to reach our goal of plus to the earnings. If we are reaching a degree, what is not for your goal but if we want to stabilise global global warming on earth and trying also to reduce the increase of other planetary limits, we need to think from that and that's around us. And when you when you think like we need to reduce by 40% of that energy consumption in 30 years, there is many things that need to be rethink. And even the way we are developing sector in the sector is based on let's see that the production and by got by organization by by users and by many by many by many people. But it makes sense in the world that make money out of that the production whether those are quality because you can any way and ends the quality by crisscrossing this kind of data sets and try to until you make something that can be useful from a marketing perspective or from that a prophetic perspective. So even we, if we create like low quality data, we, we can find ways to make its use to from a specific perspective, it is a one from the business model we are following right now. But indeed the question of thinking of as to sectors which is not based on the message that the production. Is very challenging and is very at the same time very interesting because. It's hard to think how we can get there. And also it's hard to assess what's that what what that is worth. While I'm not from a from a user perspective and we don't have constraints anymore that make make sure this is reflective process of should I keep it or not. For many people that lived in a world with lower quality or more expensive data, the process of taking pictures makes sense because there was overall constraints. Nowadays there is no the constraints of fact as well in the message that the message that the production world ecosystem is that because we are lowering constraints and making making the invisible. And and that's just my take on it. So we when when you think of the overall constraints that should be there and are not, then obviously you are producing behavior and business models that are based on the data.

00:47:41

S1: A lot of this data that is actually useful is driving the consumption. We're trying to you reduce like a lot of the biggest IT companies in the world are now advertising companies. I mean, Google is an advertising company, Facebook is an advertising company. They just the technology is a side impact of what they do. But where they make Google makes 80% of its revenue through advertising. Facebook makes 95% their advertising companies. So they're using data to actually drive superficial consumption in in the vast majority of situations. So even the so-called useful data is actually destructive of a livable planet in the future. So I think I think sooner or later we need to have a a serious policy look at data may be the secret driver of the climate crisis.

00:48:37

S2: Mm hmm. I mean, it depends because, um, depends on your definition of data. And I think we need much more precision what you call data, because between, you know, whether that's a satellite data for kind of monitoring from or from helps host data to to to advertisement to data use for the diamonds, there is a huge gap, which means that if you need to assess which that should be kept in the framework, that is a highly moral. Perspective. But how we take it the other way, we're saying like actually what's happening in Ireland is can be seen as an interesting experiment because if I remember well the agreed put the constraints on the sector on the capacity of the sector because there would be no more datacentre new datacentres in Dublin until 2028, if I remember well, which meant if that, if that production keeps increasing then it. Part of it will be what will be taken care of by the defense already existing, and some of it will be devoted elsewhere. And it will be interesting to look at the other patterns that that that we learn from this region to see how it works. But I don't know. It's not very interesting what I'm seeing at the moment. So I don't know what to think actually about what about your proposal? Okay. I need I need time to think.

00:50:18

S1: I think if you're interested in these sorts of ideas, please check out my book Worldwide Waste a Terry McArthur dot com. To hear other interesting podcasts, please visit. This is CD dot.com.

‍