Green IO #44 - Can the data center industry become circular? with Deborah Andrews
Sep 3, 2024
55:47
đź’Fueled by the ongoing artificial intelligence boom, data centers are popping up around the world like mushrooms after a good rain raising serious sustainability concerns. Hence a pressing question: can the data center industry become circular?
🎙️To get some answers, Gaël Duez welcomes a veteran in Circular Economy and Life Cycle Sustainability Assessment, Prof. Deborah Andrews, from London South Bank University, the founder and academic lead for CEDaCI.
🎙️To get some answers, Gaël Duez welcomes a veteran in Circular Economy and Life Cycle Sustainability Assessment, Prof. Deborah Andrews, from London South Bank University, the founder and academic lead for CEDaCI.
Some Takeaways:
đź’ˇ how the CEDaCI Compass tool can help data center being equipped more sustainably,
♻️ current inadequacies in recycling infrastructure for electronic waste, and
⚡ concerns about the rapid development of AI and its energy demands
❤️ Subscribe, follow, like, ... stay connected the way you want to never miss our episode, every two Tuesday!
📧 Once a month, you get carefully curated news on digital sustainability packed with exclusive Green IO contents, subscribe to the Green IO newsletter here.
đź“Ł Green IO London is on September 19th 2024 --> use the voucher GREENIOVIP to get a free ticket!
Learn more about our guest and connect:
📧 You can also send us an email at contact@greenio.com to share your feedback and suggest future guests or topics.
Deborah's sources and other references mentioned in this episode:
- CEDaCI Project
- CEDaCI Compass
- Investment in clean energy this year is set to be twice the amount going to fossil fuels
- Deforestation-free supply chains
- Life Cycle Assessment (LCA) – Everything you need to know
- Tantalum is mined in Central Africa, mainly in DRC
- Interreg NWE
- European Regional Development Fund
Transcript
Intro 00:00
There's a real need to shift thinking and business models and possibly, rather than selling it, equipment actually to sell services. So companies own equipment, so then they're responsible for maintenance and for what happens, either for extending life, which would be very much to their advantage, or for recycling, ending life.
Gael Duez 00:37
Hello everyone, welcome to Green IO with Gael Duez - that’s me! In this podcast we empower responsible technologists to build a greener digital world, one byte at a time. Twice a month on a Tuesday, our guests from across the globe share insights, tools, and alternative approaches, enabling people within the Tech sector and beyond, to boost Digital Sustainability.
And because accessible and transparent information is in the DNA of Green IO, all the references mentioned in this episode, as well as the transcript, will be in the show notes. You can find these notes on your favorite podcast platform, and, of course, on our website greenio.tech.
Can the data center industry become circular? Simple question, very complex answers.
Fueled by the ongoing artificial intelligence boom, data centers are popping up around the world like mushrooms after a good rain. For an industry which is more and more under scrutiny due to its environmental footprint, the sustainability angle cannot be overlooked anymore. Still, the main hurdle remains data, as many guests already stated in this podcast. Hence, my wish today to get insights from the program leader of an initiative which has managed to build high quality primary data on the data center industry. The CEDaCI project sounds like a Dan Brown book title, CEDaCI code to unveil all the mysterious power beneath our almighty data center industry. And to some extent, it's quite the plot. But I will let Deborah Andrews, Professor at London South Bank University and a two decades long veteran in Circular Economy and Life Cycle Sustainability Assessment reveal all of it.
Hi, Deborah. Thanks a lot for joining Green IO today.
Deborah Andrews 02:32
Thank you very much for inviting me.
Gael Duez 02:35
It's a pleasure. It's a pleasure. And I have a very direct question to ask you to kick start our discussion. Did you start the CEDaCI project out of frustration somehow?
Deborah Andrews 02:49
Okay, well, first of all, thank you for calling CEDaCI. That's the Italian interpretation. We normally describe it as CEDaCI, but hey ho, it's an acronym for the Circular Economy for the Data Center Industry. And yes, there was an element of frustration. I had worked with researchers and operators and so forth in the data center industry on a number of research projects from about 2010, 2012, and was acutely aware that the sector was very fragmented. There was the most phenomenal amount of expertise in the sector, but people worked in silos and didn't connect with each other. And consequently, there wasn't a sort of whole systems approach to the challenge of sustainability. And the industry experts in each of the sectors were doing the best that they could for their particular sub sector, but there was no consideration of the impact that those actions had in their sector. What the impact on other parts of the industry were? So it was absolutely apparent that there was a need for a whole systems approach. And it was very timely in, you know, having spoken to people who subsequently became partners in the project, that they were also acutely aware of this challenge. But being in academia, I was very lucky, being sort of slightly outside the industry, to be able to bring various representatives, stakeholders, etcetera, from the sub sectors together without having any bias.
Gael Duez 04:38
And Deborah just could you illustrate, maybe with a few examples, what silos are you referring to?
Deborah Andrews 04:45
So, one of the first things that we did in the project was to carry out a very critical appraisal of the state of the art. In other words, we did a scoping review of what was going on in the industry, and we identified eleven key sub sectors, key players within the industry, starting with suppliers and then going through design, manufacturing, etcetera. But what we did was linked the silos, or found evidence that these silos actually were related to all the different lifecycle stages of data center equipment.
Now, we focused on electrical and electronic equipment, because based on prior studies with a very extensive PhD that we ran in conjunction with HP, we found that the hotspot, if you like, the environmental hotspot, and this was looking at a whole data center. So the building or the services, M&E, etcetera. The key area of environmental impact was IT equipment, partly because of the embodied materials, the energy consumption, but also the short life of products, and in particular of servers. We found that they have ordered center equipment, and this was also reflected in a big EU report that informed lot nine that servers had the highest environmental impact. So that was the focus of the CEDaCI project. But coming back to the life cycle stages and so forth, and the various silos, then we have the installation phase. And use, of course, is incredibly important. Operational energy transport, taking stuff to and from data centers, perhaps taking to secondary market operators or recycling plants. Then we have data destruction, which could be through mechanical means, shredding, etcetera, or it could be more digital with software. And then ultimately we move on to the end of life. And we could say end of first Life, which leads to secondary market and reuse. This includes refurbishment and remanufacture. Ultimately, though, whether you send your equipment to secondary market suppliers or straight on for recycling, eventually all equipment ends up with end of life processes.
Now, ideally as much of the product should be recycled as possible. But what tends to happen is that the low hanging fruit, things like the casings and so forth, as their steel and obviously external, those recycled and the majority of the PCBs aren't, they end up in landfill. There's a real need for a shift in thinking practice to manage end of life equipment. In this CEDaCI project, one of the key things, one of our USPs, was to bring together representatives, stakeholders from these various subsectors. But one of the points that came up, because we organized co-creation workshops as part of the project, to identify what stakeholders felt they wanted, whether what they wanted was in line with what we felt would benefit them, which was a tool to aid decision making about to help with the sustainability profiling companies and so forth. But one of the USPs brought together these people, and then invariably they commented how much they had learned through the co-creation workshops, because they didn't talk to people who worked in other sub sectors. So this, again, you know, from a sort of academic or life cycle thinking perspective, you need an absolutely whole systems approach to the challenge of whatever industry you're working in to develop a circular economy. Because every action at every life cycle stage, etcetera, has caused an effect. It has an impact on all other life cycle stages.
Gael Duez 09:27
Okay, so there are so many things to unpack here from systemic thinking and your systemic approach, like the benefits of co-creation, and also what you've mentioned on e-waste. And if you, if you indulge me, I'd like to deep dive about this end of life differentiation that you just made. The first end of life, and then the second and eventually the final end of life, which is when the electronic equipment becomes e-waste. It's very important because I see today in sustainability criteria applied for RFP or bid, etcetera, or auditing the data center industry, or auditing hyperscalers, that more and more they kind of tick the box of all servers are donated to charity. All servers go to the second hand market. And so we're good, job done. I think it's a bit more complicated than that, because if the server is donated, but one year later, it ends up in a landfill somewhere in Africa, Southeast Asia, South America, or even Europe, that's not really a solution to the problem. So my question is, how important is this distinction between the first end of life and second end of life? And is it today enough taken into consideration on how to assess the sustainability of the data center industry?
Deborah Andrews 11:02
Okay, so let's think about equipment, and in particular, servers, which I've said have the highest embodied impact of data or data center equipment. And the technology, obviously, whether it's memory, processing speed, whatever, is developing incredibly rapidly. We won't even mention AI at the minute, but the sub-technical change means that an awful lot of equipment is replaced while it is still working. You know, it's functioning very, very well. Can understand, in a way, why hyperscalers and other bodies may want to have the latest, fastest, best memory, etc, equipment. But what on earth happens to the stuff that comes out of hyperscale and other centers when it's still working? Why would you want to recycle anything, take something out of service, recycle it when it still functions very well? So there's obviously a very key concern is memory and disks, be they hard drives or solid state drives. What happens to the data? We'll come back to that in a minute. But there are initiatives to encourage use of second life products, products that are still very serviceable. I mean, that's, I suppose, quite a nice analogy, in a way is, and let's assume they're all electric vehicles rather than fossil fuel driven. But, you know, maybe the hyperscalers want Ferraris, when actually many other industries would… something like Fiat Cinquecento or VW Polo will meet their transport needs. Okay, so there are drives. And certainly, I know in the UK, government and organizations like the NHS are really encouraged to use secondary market products, which one of our partners, Techbuyer, a company that is linked to them, interactive, they did a massive amount of research for their own business, but also the research fed into the CEDaCI project, they found that there were a lot of myths around the performance, particularly to do with operational energy, the performance of old, within three years old, and brand new products. If the equipment is set up correctly, then the difference in energy consumption is negligible. Obviously, it depends what compute activities you're engaged in, but like for like, energy consumption, the level is negligible. The OEMs, of course, want everybody to buy new equipment, so they're always saying, “Oh, it's better, you know, it's faster, it's more efficient, blah, blah, blah.” So coming back to the car analogy, an awful lot of public bodies in the UK, and probably in the EU as well, are being encouraged to use second life products which meet their technical needs without any problem whatsoever. But the good thing about this is the secondary market, we should have the products there, and this is a key element within the circular economy as well, of course, because it's not just about recycling at the end of life, it's about extending product life. As I said previously, why would you want to take a product out of service that is still functioning and can fulfill somebody's requirements? I think the key statement around reuse is keeping something in service for as long as technically and economically viable. That's really, really important. So the secondary market is various companies have been set up to collaborate with big hyperscale operators or smaller operators and are promoting good practice, I think, in terms of circularity and economics and resource efficiency, there are some challenges there to do with legislation and ensuring that second life products have the same warranties and so forth, and also to do with data security. Ideally, we should simply wipe drives, be they solid state or hard disk whatever, and reuse them. But there's a lot of anxiety about data security. So organizations like banks, for example, insist that drives are shredded on premise. They don't leave the bank once they're there, they come in as new products and they end up leaving as bags of tiny little bits of metal. They're shredded on site to ensure data security. But I think, and again, there are a number of really good research projects ongoing looking at data security and performance of either software wiping, whatever, so that hardware, HDDs and SDDs, SSDs can be reused.
Gael Duez 16:25
Are you optimistic about the fact that even for highly secured working environments like banks or even the military, software wiping will prevail at some point?
Deborah Andrews 16:38
I'd like to think so. I mean, I'm not an expert in this, but I would like to think so. And I think the more really robust and empirical research that can be carried out to reassure end users that software wiping or equivalent is safe as shredding, the better, really. So, yeah, fingers crossed.
Gael Duez 17:05
And Deborah, you were mentioning that there are a number of companies that have been set up to meet this new market. Could you share with us some trends? Is it still marginal, or are we witnessing a boom in the second hand market of professional IT equipment?
Deborah Andrews 17:23
It's really interesting. I think it depends very much on time, place, circumstance, etcetera. So it was interesting. During COVID for example, there was a marked increase in demand for second life products, partly because of the increase in demand for data center industry services due to homeschooling and home working, for example, but also because there were supply chain issues. And if you remember, there were all sorts of problems with chips, etcetera. So once we got over Covid and the industry really kicked off again, back to normal business as usual practices. I'll give you a little anecdote here. Okay. There was a bit of an issue around the secondary market because some manufacturers had lots of new equipment in stock before COVID but they hadn't, for various reasons sold it on.
So when there was a sort of shift in business practice generally around the world, a lot of the big manufacturers suddenly decided to flood the market with new equipment. And the price difference between that and reused equipment was negligible. So of course people wanted all purchasers, procurement teams wanted new equipment. So that had a really adverse impact on the secondary market for some companies. I'm not saying all, and it ist anecdotal, but I think it looks as though new and secondary markets are very subject to influence from external factors. You know, they're not as consistent, as stable as certainly the secondary market as we would like it to be.
Gael Duez 19:28
Did you explore with the CEDaCI project, the potential lifespan of servers if the second hand market were to be a very well functioning market? Because I've seen some colocation service providers starting to claim, and I congratulate them for this, that they keep their server for seven, sometimes eight, sometimes even nine years. But can we do better? Can we imagine a world where a server would last for 20 or 30 years? Or would it make any sense?
Deborah Andrews 20:04
First of all, if your server, and it's like, you know, your car, your Polo, your Fiat Cinquecento, whatever, even your Ferrari, if it's working well, then you should be able to keep it in service, in life or operation for a long time. The big challenge is when things, parts start to wear out or break, for whatever reason, they fail, you need to replace them. And a lot of manufacturers don't keep parts. In fact, I think legislation at the moment stipulates that manufacturers only have to supply parts for up to eight years. So if you have a ten year old server, where are you going to get the parts? It could be that you go to, say, a secondary market supplier and they have parts, but it becomes increasingly difficult over time to replace components. So that's a big challenge. The other thing is looking at changes in compute capability. We know that, for instance, there's all sorts of issues around whether you go for air cooling, to go for liquid cooling, is liquid cooling more efficient, etcetera. And AI increases the operating temperature of components. So that may well mean that we have to redesign service to manage factors like that. As compute changes, we actually need to reconfigure either layout, or if you are going for air cooling, then the number and type of fans that you have or you switch to liquid, whatever is most appropriate. So in theory, if we had modular servers that you could take out certain components of and replace with upgraded components that had common connectors and, you know, the box, the space required to house them was the same, then it seems you should be able to keep servers. You know, even if you end up with everything, all the internals, the guts of the server being replaced, the chassis, you should in theory, be able to keep reusing that for perpetuity. But whether that's feasible or not, because of changes in shift from expert liquid cooling, whatever is, we need to think about that.
Gael Duez 22:40
And before talking about the main results of the CEDaCI project, and as you mentioned, what all the stakeholders learned from each other, I've got one final clarification question. Because you were very assertive that the embodied environmental footprint within equipment is by far the biggest share of the overall environmental footprint of a data center. And if this position is not a debate at all when it comes to end user equipment or devices at home, whether it's laptop, smartphone, etcetera, etcetera, I heard some different opinions where it's more like 50-50, because professional equipment lasts longer. Some people advocate that actually the use phase, and especially the energy consumption during the use phase is far from being negligible. And for some of them, it's even the majority of the environmental footprint. So maybe could you clarify whether you were mentioning only the carbon or other environmental impacts, or even maybe the carbon. Your calculations make it clear that embodied carbon is even bigger than energy consumption, GHG emission. I was looking for a bit of a clarification here where you stand on this debate.
Deborah Andrews 23:59
Okay. So when we said that in our calculations that the largest impact in the data center was the IT equipment, when we said that the largest impact was the IT equipment, we based our model on a data center that lasted for 60 years. The actual infrastructure, the building, some of the M&E, was replaced after 20 years, etcetera. But the IT product life was based on three to five years, okay? And that's for hyperscalers, that's a long, long time. Although Google now says that they're going to keep servers in life for five years, we'll see about that.
So coming back to... So i if you decide to pull your data center down after 20 years and build a new one, then that ratio of impact from IT to building will change. But just be clear about that, okay. The second thing to be really, really clear about is the kind of metrics you use when you are looking at impact. So carbon, obviously, and carbon equivalent, it could be methane or other hydrocarbons. Carbon is only one metric, carbon and carbon equivalents. And you exclude a whole array in fact, thousands of other impacts and outputs, inputs, etc. And impacts, when you only look at carbon and carbon equivalent. So you're not thinking about the impact of water, the impact of gold mining, for example, which is incredibly toxic, the tailings can be, may have one of the highest environmental impacts, etcetera. And the impact is not just environmental impact, it can be very detrimental to ecosystems. So toxic substances, mercury, arsenic, etcetera, used in mining processes will obviously have an impact on people living in the area if they get into the water supply and into soil and so forth. So if you're looking at carbon, you exclude all of those factors. I'm not 100% against carbon assessment. I just think we have to think about it in relation to everything else. It was really, it was the original sort of metric linked to life cycle assessment dating from the 1960s. So when LCA first began, it just considered energy, be it operational or embodied. And that's where the linked carbon assessment comes from. It's incredibly important when we're thinking about climate change. We can't underestimate its significance. However, we do need to think about all the other impacts as well. So we carried out some studies of exactly the same piece of equipment, one looking at carbon, operational and embodied carbon, and the other looking at a comprehensive life cycle assessment, which looked at and included thousands of inputs and outputs. But just on a carbon study, we found that operational energy, when you're looking at lifespan, let's say five years, operational energy accounted for 85% of impact, or the carbon in the operational energy, 85% of impact, and embodied in energy was only 15. Okay, so with the newer equipment, that ratio shifted a bit and the sort of makeup shifted to 20% for embodied impact, and 80% for operational impact over a five year life. So it's no wonder that when the data center industry is being guided by, you know, the need to reduce carbon, be it embodied, operational, it's no wonder that you focus, the industry focused on operational energy initially, and also I think it's easier to manage to make change. So now we see, you know, more use of renewables, etcetera. And also improved operating efficiency of equipment itself.
When we look at comprehensive life cycle assessment, we've got a couple of surprises. We didn't think that the difference between carbon and full LCA would be so significant. So the same piece of equipment or two pieces of equipment. The first was, as I said, a really old server, and the ratio of operational to embodied impact was about 80:20 for carbon assessment, and again, it was lower. It was, I can't remember something like 75:25. But what, this is the really, really, really key finding. When we looked at the comparatively new piece of equipment, which was from about 2017, we found that the embodied and operational impacts were about the same. It was about a 50:50 split. So that was really, that means, that was a real surprise. And it really highlights the necessity of examining, measuring, monitoring, you know, improving the physical resource efficiency, increasing use of recycled materials, increasing recycling processes, and building a decent infrastructure to do that, changing practice with things like encouraging, if you can't reuse a whole piece of equipment, can we harvest components and reuse a individual components, etcetera. So that was a very revelationary, very revealing study.
Gael Duez 30:15
This shift from 80:20 to 50:50 was mostly due to energy savings, energy optimization of newer equipment, or was it also, as you said, because of lower cost of building equipment, thanks to the use of recycling materials, etcetera, etcetera.
Deborah Andrews 30:36
Okay, so this 50:50 split, it was a lot to do with improved operational energy. When you're looking at the full life cycle assessment, you are considering things like what happens during mining processes. It's not just the energy or the toxicity, etcetera, and also things like what percentage. Well, now we should be thinking about the percentage of recycled materials included, that, in theory, should reduce environmental impact.
Gael Duez 31:14
So now that we laid the ground for a better understanding of all this environmental footprint, etcetera, let's go back to the actual findings of the CEDaCI project, both for the stakeholders, but also for an average, I would say, data center operator. What are the findings he or she should be aware of? And what about this tool that you developed called the compass?
Deborah Andrews 31:38
The CEDaCI Compass, the circular data center compass. That was a key output from the CEDaCI project. And the majority of the research that we did underpinned the development of this digital tool. It's free to use. You can find it, access it via the project website at cedaci.org. The idea was that we wanted to help people working in the industry to make informed decisions about how to support their transition to sustainability and circularity. And one of the things that we are very keen to do well, apart from being absolutely objective, it's completely non judgmental. We don't offer any, you know, this is right, this is wrong. We just present the results. One of the things that we did as well was to separate out the three sorts of tenets or pillars of sustainability. So you put your information into the tool, or you select various criteria, and then when you see the results, comparing two different servers, you can compare the environmental impact, the social impact, the economic impact, so that, you know, it could be, as an operator, you're more concerned about social factors than environmental, so you base your decision on that. Or I would imagine most people think about economics as their priority driver. But, you know, being aware of the other criteria is really important. If you lump together, you know, these three key tenets of sustainability, you get slightly inaccurate results, you know, you don't know what I mean. It could be the social impact of one server is very high, but its environmental, adverse environmental impact is very low, or the other way around. So that's why we separated those points out. The other thing that we included was a criticality indicator. Now the EU and, well, UK to a certain extent, but certainly the EU has become increasingly conscious of resource efficiency and has identified, now, 30 materials that are what they define as critical. This is because of the amount of resource that is as yet unmined in the surface, the amount of material that is currently recycled. And I would like to include as well the possibility of substitution. But the other really, really important is the geopolitical factors, where on earth is the material located? Because that has a very significant impact on availability. So the critical raw materials are basically defined as those that are of major technical and economic significance to the EU and UK. So we included a criticality indicator to raise awareness of these materials. The use of these materials in data center equipment, all electronic equipment, uses some critical raw materials of some type. We cannot make, for example, mobile phones without tantalum, which is essential for capacitors. Tantalum is mined in Central Africa, mainly in DRC, Congo, where mining practices are eye wateringly horrific. They are environmentally and socially damaging. So we wanted to raise awareness of those sort of issues as well, to encourage better practice. So the compass was developed. As I said, it's a free to access online tool. And it was basically to inform potential end users about the impact of their choices. The other thing that was really, really important at the moment, we don't have anything like the scale of recycling infrastructure that we need to manage all the equipment that's currently in circulation, let alone all the electrical and electronic equipment that will come into the waste stream imminently. We've got a huge problem with this. There's a collection globally and it does depend where you are based globally and as does reuse. Incidentally, people with less expendable income tend to be more frugal by necessity rather than intent, I think or wish. Overall, the collection rate of e-waste is less than 20% globally. And that includes consumer as well as commercial products. But we don't have anything like the infrastructure that we need to recycle this anyway. The recycling processes at the moment, they focus on anything with iron in it. So steel, copper, aluminum, and gold. And there's not very much gold, you know, by mass, it's comparatively little gold in electronic equipment. But of course, its inherent economic value makes it attractive to recycle it. So there are masses and masses of materials, many of which are critical on the critical raw materials list, which aren't recycled. Unless we get our act together and develop a proper recycling infrastructure, there's significant potential for disruption to supply chains.
Gael Duez 38:54
But do you believe it is possible? My understanding of the chipmaking industry, or even slightly less complicated part of the IT industry, the design itself is so complicated, is so… the different metals are melted together to create alloy. You've got ceramic, etcetera, that I don't even know if it's feasible to recycle the way I would say the average John do understands it, which is we will extract to reuse it the same way. And it's more down cycling or it's even. I mean, I honestly wonder if recycling is really the way forward. So don't get me wrong, instead of being able to reuse for super long period of time, as you previously mentioned, having components that are interoperable on open standard etcetera, and saying, okay, you know, this memory card might be 20 years old, so it's ridiculous the amount of data you've got. But hey, I've still got half a billion of them. And if I've put them in some racks, it's still a decent, you know, decent enough, or whatever. But my point is, I think we are fighting an uphill battle if we really want to recycle, like extracting the tantalum you were mentioning, or the cobalt or whatever, rather than redesigning our industrial process and making also sure that the warranty period is so big that so long, that actually we shift the burden of recycling to producer, which are eventually, ultimately responsible for putting things on the market that are absolutely not recyclable. And that would end up being e-waste in a matter of years rather than decades. And when you see the environmental footprint of everything that you describe, we should talk in decades rather than in years. But that's a personal opinion. Sorry, but my point. What do you think about the feasibility of recycling? Or are we talking about a slightly different approach in the recycling industry for the IT equipment?
Deborah Andrews 41:05
I think we need to have all of those things, really. I don't think there's not one size fits all solution. I think there are some really massive challenges with electronic components, for example, because of the way that they function. You're looking at atomic levels, the way that if you're creating signals, etcetera, the way that atoms and the subatomic particles behave. So that is obviously going to limit the way in which components are designed and manufactured at the moment. Maybe in the future they will discover different approaches to data transfer or signaling, switching, whatever it may be. I don't know too much about quantum, and I don't know how this is going to change things, if at all, if we exchange one set of problems for another. I don't know, but I'd certainly like to find out an awful lot more about it.
But coming back to your question about end of life and so forth, I absolutely agree that we need a different approach to the manufacture of many products. I think we need to really focus on the things that can be upgraded, swapped, repaired, etcetera. Again, most or many electronic components, it's nothing. You know, they're so tiny, it's impossible to repair them. So focus on the stuff that we can repair and upgrade and keep in service for as long as possible. The other stuff, we need to certainly develop better recycling capability. But there's an argument that's put forward that some of the materials, the economic value is fine if you have a kilo of stuff, but by mass per component, the mass in individual components and in servers as a whole is very low. So the economic value of any particular materials reclaimed from the server will be low. So we need critical mass to make development of recycling and reclamation technologies for particular materials economically viable. That's going to be the driver. The other big challenge we've got with recycling is sometimes the processes, and we're not. You know, there are lots of ways of doing this. One of the partners in this CEDaCI project, a company called TND or Terra Nova Developments, they developed some new recycling processes to reclaim materials that aren't commonly reclaimed. And they use a mix of thermal and chemical processes. And because you're using more than one process, of course, that increases cost. But, you know, if you have critical mass of stuff coming through the system, it does become economically viable eventually.
Deborah Andrews 44:23
One of the big challenges is very often, but if you have a printed circuit board with a huge number of materials, embodied materials, you're processing to reclaim one or two or three, those processes can damage other materials, and so you can't reclaim those. So, you know, I think at the moment, it's impossible to reclaim all of the materials in PCBs, which are the biggest challenge for recycling. So I think we need to think about when we're designing, not just designing for here now and in use to design, thinking about how can these things be easily disassembled to facilitate, if not, you know, chucking the material into a smelter to recycle it, but actually being able to reuse the materials in the… as soon as we take them off one product, we can put them on another.
Gael Duez 45:22
Do IT equipment manufacturers today start to tackle the issue? Or are they still mostly in the business as usual approach and not at all incorporating, as you said, a repairability aspect when the design sinks, even a bit of a recycling aspect?
Deborah Andrews 45:45
I think there's, and I'm not going to highlight particular companies, but I think there's quite a lot of smoke and mirrors and greenwashing. And this, again, it's anecdotal from personal experience of, you know, I worked through this CEDaCI project. They all still want to sell new equipment, and certain companies say that they have a kind of closed loop, but it's a very open closed loop, shall we say? They're not responsible. They do take equipment, but then they sell it onto secondary market agencies. They're no longer responsible for that equipment once they've sold it on. There's a real need to shift thinking and business models and possibly, rather than selling IT equipment actually to sell services. So, you know, you rent, I don't know whether you do it by compute capability or operating time or whatever. So then they're responsible for maintenance and for what happens, either for extending life, which would be very much to their advantage, or for recycling at the end of life.
Gael Duez 47:00
Which is what is slightly happening with the big three hyperscalers because they've started designing their own servers. And I could bet that they thought about the fact that even from a financial perspective, the longer you keep this equipment, which is on your own cost base, the better it is for your bottom line. And I have a final question for you, because you're privileged witness of the data center industry for almost a decade, two decades almost now, and there is a bit of an elephant in the room that you actually, you teased us several times during the interview about AI and more generally, I would say, about the trends. And I would like to share with you an anecdote while I was recording this very enlightening episode with Professor PS Lee in Singapore, who's one of the best experts in the water cooling techniques and more generally on building energy efficient data centers, especially in tropical climates. We had a really fruitful discussion, and he's a big advocate of technical optimization, and he knows a lot about these topics, etcetera. And at some point in our discussion, he posed, and he had this kind of overwhelmed moment, you know, when you've got just too much weight on your shoulders. He was like, but you know, Gael, at some point we will have to ask also the question of the level of consumption of compute that we want in our society. Because despite all the efforts that I'm doing to reduce energy consumption, the current trends, and especially the AI boom, I don't see how I can make it. And it was this kind of face, like, I simply don't know how I will be able to manage such an exponential consumption in energy, even if overnight all data centers in the world would switch to water cooling, super efficient water cooling techniques and whatnot. And it kind of struck me like even someone within the tech industry and such a strong advocate of technical optimization saying, wow, but the trend is really worrisome. Is it something that you're aligned on, or are you a bit more optimistic about the current trend in the data center industry?
Deborah Andrews 49:29
I think it's really scary. Forget about the ethics and what AI can and can't do. Forget about that. I think it's really scary because it seems that the industry is racing, racing forwards to develop either new data centers, new equipment to manage AI, etcetera. But my feeling is that some of the, what we've learned, good practice, etcetera, is forgotten in that race. So it's a business as usual approach. Build, install, run, replace, run. It's very worrying because the physical resources and energy required to manage AI operate, even the simplest operation, the demand for energy, is astronomical. So there are arguments saying, oh, well, AI is going to measure this, that and the other, and improve, you know, this, that and the other, but, or improve the operational efficiency of this, that and the other. But I don't think we've done any kind of calculations at all to see whether the benefits of running an AI software or, you know, program activity, whatever, to assess resource efficiency, whether the benefits achieved through the resource efficiency or more significant than the impact of running the AI operation. We need to really, and it's not going to happen, but it would be great to pause for a year and just to sort of examine some of these factors and to see where the benefits of AI really lie. So if we do carry on business as usual, and I know there are various regulations coming in from, you know, digital sustainability and so forth in the EU, but I don't think they're going to have a massive impact on, certainly the speed of development of AI is far faster than the implementation of these new regulations. But the other thing that's worrying, if there are constraints of operating in the EU, for example, does it simply mean that providers will go elsewhere, they'll build in locations where the regulations don't apply. And so we move our problems to another part of the world. In effect, you know, more buildings in Africa, Asia, South America.
Gael Duez 52:05
And from a cold financial perspective, do you think that the current trend of building data centers everywhere in increasing compute capacity is sustainable? Or do you foresee some bottleneck or even some forced pose because of resource exhaustion?
Deborah Andrews 52:25
There may be a quantity crunch, unless we think about extending product life, as we've already mentioned, and recycling, reclaiming more materials, changing business practice, business thinking. I think that there's potential for quantity crunch, but whether operators will start to charge for access to digital services. I mean, at the moment, you know, you buy your phone, you have a package, and you can contact anywhere in the world whenever you want. Okay, you pay for apps and so forth, but actually you're not really paying very much for the digital services that enable those apps to function. And a lot of things are free anyway. So will we have to pay for digital services? Is that one way of monitoring or constraining digital activity or not? That's one question. But the other thing is, is it ethical? You know, if we think we're in a luxurious position in Western Europe in terms of economics, although there are people in digital poverty, but, you know, generally, as a child or a university student, if you don't have computing equipment at home, you can go to your study institution and access digital technology there. But that's not the case in many populations in developing countries where digital tech is a luxury for the upper echelons of society. And yet we can see how access to a phone, not smartphone, just an ordinary old fashioned phone, has empowered women, for example, helping them to set up businesses and so forth. Is it right to charge them for data or do we make data charging… is it sort of socially stratified or according to income bracket? I don't know. I think we need to be a lot more visionary, look forwards and be proactive and anticipate problems and design them out before they happen. Now, whether that's… I think that's possible with some equipment, but whether it's possible for, you know, in terms of human behaviour and so forth, there's a whole other matter altogether.
Gael Duez 54:18
It makes total sense. Thanks a lot, Deborah, for joining. That was very enlightening and a unique perspective on the data center industry, environmental footprint and what could be done to reduce it. So thanks a lot. Once again, it was a pleasure to see you and hope to see you to Green IO London as well.
Deborah Andrews 54:36
Thank you very much. It's been an absolute pleasure. Thank you.
Outro 54:42
Thank you for listening to this Green IO episode. If you enjoyed it, please share it via email, on LinkedIn, on Facebook, on Twitter, if you are still there. We are an independent media and word of mouth is the only way to get more listeners. I don't ask you to rate it five stars on Spotify or Apple Podcasts because of course you already did it, didnt you?
It's time for you to grab a book or enjoy a good article. And guess what? You can find many ideas in the latest edition of the Green IO monthly newsletter. And don't forget to book your ticket for the next Green IO conference in London on September 19. Good news for your holiday budget. You can still get a free ticket using the voucher GREENIOVIP. Just make sure to have one before they're all gone. I'm looking forward to meeting you back in six weeks to help you, fellow responsible technologists build a queener digital world.
Roxanne 55:40
One byte at a time.
There's a real need to shift thinking and business models and possibly, rather than selling it, equipment actually to sell services. So companies own equipment, so then they're responsible for maintenance and for what happens, either for extending life, which would be very much to their advantage, or for recycling, ending life.
Gael Duez 00:37
Hello everyone, welcome to Green IO with Gael Duez - that’s me! In this podcast we empower responsible technologists to build a greener digital world, one byte at a time. Twice a month on a Tuesday, our guests from across the globe share insights, tools, and alternative approaches, enabling people within the Tech sector and beyond, to boost Digital Sustainability.
And because accessible and transparent information is in the DNA of Green IO, all the references mentioned in this episode, as well as the transcript, will be in the show notes. You can find these notes on your favorite podcast platform, and, of course, on our website greenio.tech.
Can the data center industry become circular? Simple question, very complex answers.
Fueled by the ongoing artificial intelligence boom, data centers are popping up around the world like mushrooms after a good rain. For an industry which is more and more under scrutiny due to its environmental footprint, the sustainability angle cannot be overlooked anymore. Still, the main hurdle remains data, as many guests already stated in this podcast. Hence, my wish today to get insights from the program leader of an initiative which has managed to build high quality primary data on the data center industry. The CEDaCI project sounds like a Dan Brown book title, CEDaCI code to unveil all the mysterious power beneath our almighty data center industry. And to some extent, it's quite the plot. But I will let Deborah Andrews, Professor at London South Bank University and a two decades long veteran in Circular Economy and Life Cycle Sustainability Assessment reveal all of it.
Hi, Deborah. Thanks a lot for joining Green IO today.
Deborah Andrews 02:32
Thank you very much for inviting me.
Gael Duez 02:35
It's a pleasure. It's a pleasure. And I have a very direct question to ask you to kick start our discussion. Did you start the CEDaCI project out of frustration somehow?
Deborah Andrews 02:49
Okay, well, first of all, thank you for calling CEDaCI. That's the Italian interpretation. We normally describe it as CEDaCI, but hey ho, it's an acronym for the Circular Economy for the Data Center Industry. And yes, there was an element of frustration. I had worked with researchers and operators and so forth in the data center industry on a number of research projects from about 2010, 2012, and was acutely aware that the sector was very fragmented. There was the most phenomenal amount of expertise in the sector, but people worked in silos and didn't connect with each other. And consequently, there wasn't a sort of whole systems approach to the challenge of sustainability. And the industry experts in each of the sectors were doing the best that they could for their particular sub sector, but there was no consideration of the impact that those actions had in their sector. What the impact on other parts of the industry were? So it was absolutely apparent that there was a need for a whole systems approach. And it was very timely in, you know, having spoken to people who subsequently became partners in the project, that they were also acutely aware of this challenge. But being in academia, I was very lucky, being sort of slightly outside the industry, to be able to bring various representatives, stakeholders, etcetera, from the sub sectors together without having any bias.
Gael Duez 04:38
And Deborah just could you illustrate, maybe with a few examples, what silos are you referring to?
Deborah Andrews 04:45
So, one of the first things that we did in the project was to carry out a very critical appraisal of the state of the art. In other words, we did a scoping review of what was going on in the industry, and we identified eleven key sub sectors, key players within the industry, starting with suppliers and then going through design, manufacturing, etcetera. But what we did was linked the silos, or found evidence that these silos actually were related to all the different lifecycle stages of data center equipment.
Now, we focused on electrical and electronic equipment, because based on prior studies with a very extensive PhD that we ran in conjunction with HP, we found that the hotspot, if you like, the environmental hotspot, and this was looking at a whole data center. So the building or the services, M&E, etcetera. The key area of environmental impact was IT equipment, partly because of the embodied materials, the energy consumption, but also the short life of products, and in particular of servers. We found that they have ordered center equipment, and this was also reflected in a big EU report that informed lot nine that servers had the highest environmental impact. So that was the focus of the CEDaCI project. But coming back to the life cycle stages and so forth, and the various silos, then we have the installation phase. And use, of course, is incredibly important. Operational energy transport, taking stuff to and from data centers, perhaps taking to secondary market operators or recycling plants. Then we have data destruction, which could be through mechanical means, shredding, etcetera, or it could be more digital with software. And then ultimately we move on to the end of life. And we could say end of first Life, which leads to secondary market and reuse. This includes refurbishment and remanufacture. Ultimately, though, whether you send your equipment to secondary market suppliers or straight on for recycling, eventually all equipment ends up with end of life processes.
Now, ideally as much of the product should be recycled as possible. But what tends to happen is that the low hanging fruit, things like the casings and so forth, as their steel and obviously external, those recycled and the majority of the PCBs aren't, they end up in landfill. There's a real need for a shift in thinking practice to manage end of life equipment. In this CEDaCI project, one of the key things, one of our USPs, was to bring together representatives, stakeholders from these various subsectors. But one of the points that came up, because we organized co-creation workshops as part of the project, to identify what stakeholders felt they wanted, whether what they wanted was in line with what we felt would benefit them, which was a tool to aid decision making about to help with the sustainability profiling companies and so forth. But one of the USPs brought together these people, and then invariably they commented how much they had learned through the co-creation workshops, because they didn't talk to people who worked in other sub sectors. So this, again, you know, from a sort of academic or life cycle thinking perspective, you need an absolutely whole systems approach to the challenge of whatever industry you're working in to develop a circular economy. Because every action at every life cycle stage, etcetera, has caused an effect. It has an impact on all other life cycle stages.
Gael Duez 09:27
Okay, so there are so many things to unpack here from systemic thinking and your systemic approach, like the benefits of co-creation, and also what you've mentioned on e-waste. And if you, if you indulge me, I'd like to deep dive about this end of life differentiation that you just made. The first end of life, and then the second and eventually the final end of life, which is when the electronic equipment becomes e-waste. It's very important because I see today in sustainability criteria applied for RFP or bid, etcetera, or auditing the data center industry, or auditing hyperscalers, that more and more they kind of tick the box of all servers are donated to charity. All servers go to the second hand market. And so we're good, job done. I think it's a bit more complicated than that, because if the server is donated, but one year later, it ends up in a landfill somewhere in Africa, Southeast Asia, South America, or even Europe, that's not really a solution to the problem. So my question is, how important is this distinction between the first end of life and second end of life? And is it today enough taken into consideration on how to assess the sustainability of the data center industry?
Deborah Andrews 11:02
Okay, so let's think about equipment, and in particular, servers, which I've said have the highest embodied impact of data or data center equipment. And the technology, obviously, whether it's memory, processing speed, whatever, is developing incredibly rapidly. We won't even mention AI at the minute, but the sub-technical change means that an awful lot of equipment is replaced while it is still working. You know, it's functioning very, very well. Can understand, in a way, why hyperscalers and other bodies may want to have the latest, fastest, best memory, etc, equipment. But what on earth happens to the stuff that comes out of hyperscale and other centers when it's still working? Why would you want to recycle anything, take something out of service, recycle it when it still functions very well? So there's obviously a very key concern is memory and disks, be they hard drives or solid state drives. What happens to the data? We'll come back to that in a minute. But there are initiatives to encourage use of second life products, products that are still very serviceable. I mean, that's, I suppose, quite a nice analogy, in a way is, and let's assume they're all electric vehicles rather than fossil fuel driven. But, you know, maybe the hyperscalers want Ferraris, when actually many other industries would… something like Fiat Cinquecento or VW Polo will meet their transport needs. Okay, so there are drives. And certainly, I know in the UK, government and organizations like the NHS are really encouraged to use secondary market products, which one of our partners, Techbuyer, a company that is linked to them, interactive, they did a massive amount of research for their own business, but also the research fed into the CEDaCI project, they found that there were a lot of myths around the performance, particularly to do with operational energy, the performance of old, within three years old, and brand new products. If the equipment is set up correctly, then the difference in energy consumption is negligible. Obviously, it depends what compute activities you're engaged in, but like for like, energy consumption, the level is negligible. The OEMs, of course, want everybody to buy new equipment, so they're always saying, “Oh, it's better, you know, it's faster, it's more efficient, blah, blah, blah.” So coming back to the car analogy, an awful lot of public bodies in the UK, and probably in the EU as well, are being encouraged to use second life products which meet their technical needs without any problem whatsoever. But the good thing about this is the secondary market, we should have the products there, and this is a key element within the circular economy as well, of course, because it's not just about recycling at the end of life, it's about extending product life. As I said previously, why would you want to take a product out of service that is still functioning and can fulfill somebody's requirements? I think the key statement around reuse is keeping something in service for as long as technically and economically viable. That's really, really important. So the secondary market is various companies have been set up to collaborate with big hyperscale operators or smaller operators and are promoting good practice, I think, in terms of circularity and economics and resource efficiency, there are some challenges there to do with legislation and ensuring that second life products have the same warranties and so forth, and also to do with data security. Ideally, we should simply wipe drives, be they solid state or hard disk whatever, and reuse them. But there's a lot of anxiety about data security. So organizations like banks, for example, insist that drives are shredded on premise. They don't leave the bank once they're there, they come in as new products and they end up leaving as bags of tiny little bits of metal. They're shredded on site to ensure data security. But I think, and again, there are a number of really good research projects ongoing looking at data security and performance of either software wiping, whatever, so that hardware, HDDs and SDDs, SSDs can be reused.
Gael Duez 16:25
Are you optimistic about the fact that even for highly secured working environments like banks or even the military, software wiping will prevail at some point?
Deborah Andrews 16:38
I'd like to think so. I mean, I'm not an expert in this, but I would like to think so. And I think the more really robust and empirical research that can be carried out to reassure end users that software wiping or equivalent is safe as shredding, the better, really. So, yeah, fingers crossed.
Gael Duez 17:05
And Deborah, you were mentioning that there are a number of companies that have been set up to meet this new market. Could you share with us some trends? Is it still marginal, or are we witnessing a boom in the second hand market of professional IT equipment?
Deborah Andrews 17:23
It's really interesting. I think it depends very much on time, place, circumstance, etcetera. So it was interesting. During COVID for example, there was a marked increase in demand for second life products, partly because of the increase in demand for data center industry services due to homeschooling and home working, for example, but also because there were supply chain issues. And if you remember, there were all sorts of problems with chips, etcetera. So once we got over Covid and the industry really kicked off again, back to normal business as usual practices. I'll give you a little anecdote here. Okay. There was a bit of an issue around the secondary market because some manufacturers had lots of new equipment in stock before COVID but they hadn't, for various reasons sold it on.
So when there was a sort of shift in business practice generally around the world, a lot of the big manufacturers suddenly decided to flood the market with new equipment. And the price difference between that and reused equipment was negligible. So of course people wanted all purchasers, procurement teams wanted new equipment. So that had a really adverse impact on the secondary market for some companies. I'm not saying all, and it ist anecdotal, but I think it looks as though new and secondary markets are very subject to influence from external factors. You know, they're not as consistent, as stable as certainly the secondary market as we would like it to be.
Gael Duez 19:28
Did you explore with the CEDaCI project, the potential lifespan of servers if the second hand market were to be a very well functioning market? Because I've seen some colocation service providers starting to claim, and I congratulate them for this, that they keep their server for seven, sometimes eight, sometimes even nine years. But can we do better? Can we imagine a world where a server would last for 20 or 30 years? Or would it make any sense?
Deborah Andrews 20:04
First of all, if your server, and it's like, you know, your car, your Polo, your Fiat Cinquecento, whatever, even your Ferrari, if it's working well, then you should be able to keep it in service, in life or operation for a long time. The big challenge is when things, parts start to wear out or break, for whatever reason, they fail, you need to replace them. And a lot of manufacturers don't keep parts. In fact, I think legislation at the moment stipulates that manufacturers only have to supply parts for up to eight years. So if you have a ten year old server, where are you going to get the parts? It could be that you go to, say, a secondary market supplier and they have parts, but it becomes increasingly difficult over time to replace components. So that's a big challenge. The other thing is looking at changes in compute capability. We know that, for instance, there's all sorts of issues around whether you go for air cooling, to go for liquid cooling, is liquid cooling more efficient, etcetera. And AI increases the operating temperature of components. So that may well mean that we have to redesign service to manage factors like that. As compute changes, we actually need to reconfigure either layout, or if you are going for air cooling, then the number and type of fans that you have or you switch to liquid, whatever is most appropriate. So in theory, if we had modular servers that you could take out certain components of and replace with upgraded components that had common connectors and, you know, the box, the space required to house them was the same, then it seems you should be able to keep servers. You know, even if you end up with everything, all the internals, the guts of the server being replaced, the chassis, you should in theory, be able to keep reusing that for perpetuity. But whether that's feasible or not, because of changes in shift from expert liquid cooling, whatever is, we need to think about that.
Gael Duez 22:40
And before talking about the main results of the CEDaCI project, and as you mentioned, what all the stakeholders learned from each other, I've got one final clarification question. Because you were very assertive that the embodied environmental footprint within equipment is by far the biggest share of the overall environmental footprint of a data center. And if this position is not a debate at all when it comes to end user equipment or devices at home, whether it's laptop, smartphone, etcetera, etcetera, I heard some different opinions where it's more like 50-50, because professional equipment lasts longer. Some people advocate that actually the use phase, and especially the energy consumption during the use phase is far from being negligible. And for some of them, it's even the majority of the environmental footprint. So maybe could you clarify whether you were mentioning only the carbon or other environmental impacts, or even maybe the carbon. Your calculations make it clear that embodied carbon is even bigger than energy consumption, GHG emission. I was looking for a bit of a clarification here where you stand on this debate.
Deborah Andrews 23:59
Okay. So when we said that in our calculations that the largest impact in the data center was the IT equipment, when we said that the largest impact was the IT equipment, we based our model on a data center that lasted for 60 years. The actual infrastructure, the building, some of the M&E, was replaced after 20 years, etcetera. But the IT product life was based on three to five years, okay? And that's for hyperscalers, that's a long, long time. Although Google now says that they're going to keep servers in life for five years, we'll see about that.
So coming back to... So i if you decide to pull your data center down after 20 years and build a new one, then that ratio of impact from IT to building will change. But just be clear about that, okay. The second thing to be really, really clear about is the kind of metrics you use when you are looking at impact. So carbon, obviously, and carbon equivalent, it could be methane or other hydrocarbons. Carbon is only one metric, carbon and carbon equivalents. And you exclude a whole array in fact, thousands of other impacts and outputs, inputs, etc. And impacts, when you only look at carbon and carbon equivalent. So you're not thinking about the impact of water, the impact of gold mining, for example, which is incredibly toxic, the tailings can be, may have one of the highest environmental impacts, etcetera. And the impact is not just environmental impact, it can be very detrimental to ecosystems. So toxic substances, mercury, arsenic, etcetera, used in mining processes will obviously have an impact on people living in the area if they get into the water supply and into soil and so forth. So if you're looking at carbon, you exclude all of those factors. I'm not 100% against carbon assessment. I just think we have to think about it in relation to everything else. It was really, it was the original sort of metric linked to life cycle assessment dating from the 1960s. So when LCA first began, it just considered energy, be it operational or embodied. And that's where the linked carbon assessment comes from. It's incredibly important when we're thinking about climate change. We can't underestimate its significance. However, we do need to think about all the other impacts as well. So we carried out some studies of exactly the same piece of equipment, one looking at carbon, operational and embodied carbon, and the other looking at a comprehensive life cycle assessment, which looked at and included thousands of inputs and outputs. But just on a carbon study, we found that operational energy, when you're looking at lifespan, let's say five years, operational energy accounted for 85% of impact, or the carbon in the operational energy, 85% of impact, and embodied in energy was only 15. Okay, so with the newer equipment, that ratio shifted a bit and the sort of makeup shifted to 20% for embodied impact, and 80% for operational impact over a five year life. So it's no wonder that when the data center industry is being guided by, you know, the need to reduce carbon, be it embodied, operational, it's no wonder that you focus, the industry focused on operational energy initially, and also I think it's easier to manage to make change. So now we see, you know, more use of renewables, etcetera. And also improved operating efficiency of equipment itself.
When we look at comprehensive life cycle assessment, we've got a couple of surprises. We didn't think that the difference between carbon and full LCA would be so significant. So the same piece of equipment or two pieces of equipment. The first was, as I said, a really old server, and the ratio of operational to embodied impact was about 80:20 for carbon assessment, and again, it was lower. It was, I can't remember something like 75:25. But what, this is the really, really, really key finding. When we looked at the comparatively new piece of equipment, which was from about 2017, we found that the embodied and operational impacts were about the same. It was about a 50:50 split. So that was really, that means, that was a real surprise. And it really highlights the necessity of examining, measuring, monitoring, you know, improving the physical resource efficiency, increasing use of recycled materials, increasing recycling processes, and building a decent infrastructure to do that, changing practice with things like encouraging, if you can't reuse a whole piece of equipment, can we harvest components and reuse a individual components, etcetera. So that was a very revelationary, very revealing study.
Gael Duez 30:15
This shift from 80:20 to 50:50 was mostly due to energy savings, energy optimization of newer equipment, or was it also, as you said, because of lower cost of building equipment, thanks to the use of recycling materials, etcetera, etcetera.
Deborah Andrews 30:36
Okay, so this 50:50 split, it was a lot to do with improved operational energy. When you're looking at the full life cycle assessment, you are considering things like what happens during mining processes. It's not just the energy or the toxicity, etcetera, and also things like what percentage. Well, now we should be thinking about the percentage of recycled materials included, that, in theory, should reduce environmental impact.
Gael Duez 31:14
So now that we laid the ground for a better understanding of all this environmental footprint, etcetera, let's go back to the actual findings of the CEDaCI project, both for the stakeholders, but also for an average, I would say, data center operator. What are the findings he or she should be aware of? And what about this tool that you developed called the compass?
Deborah Andrews 31:38
The CEDaCI Compass, the circular data center compass. That was a key output from the CEDaCI project. And the majority of the research that we did underpinned the development of this digital tool. It's free to use. You can find it, access it via the project website at cedaci.org. The idea was that we wanted to help people working in the industry to make informed decisions about how to support their transition to sustainability and circularity. And one of the things that we are very keen to do well, apart from being absolutely objective, it's completely non judgmental. We don't offer any, you know, this is right, this is wrong. We just present the results. One of the things that we did as well was to separate out the three sorts of tenets or pillars of sustainability. So you put your information into the tool, or you select various criteria, and then when you see the results, comparing two different servers, you can compare the environmental impact, the social impact, the economic impact, so that, you know, it could be, as an operator, you're more concerned about social factors than environmental, so you base your decision on that. Or I would imagine most people think about economics as their priority driver. But, you know, being aware of the other criteria is really important. If you lump together, you know, these three key tenets of sustainability, you get slightly inaccurate results, you know, you don't know what I mean. It could be the social impact of one server is very high, but its environmental, adverse environmental impact is very low, or the other way around. So that's why we separated those points out. The other thing that we included was a criticality indicator. Now the EU and, well, UK to a certain extent, but certainly the EU has become increasingly conscious of resource efficiency and has identified, now, 30 materials that are what they define as critical. This is because of the amount of resource that is as yet unmined in the surface, the amount of material that is currently recycled. And I would like to include as well the possibility of substitution. But the other really, really important is the geopolitical factors, where on earth is the material located? Because that has a very significant impact on availability. So the critical raw materials are basically defined as those that are of major technical and economic significance to the EU and UK. So we included a criticality indicator to raise awareness of these materials. The use of these materials in data center equipment, all electronic equipment, uses some critical raw materials of some type. We cannot make, for example, mobile phones without tantalum, which is essential for capacitors. Tantalum is mined in Central Africa, mainly in DRC, Congo, where mining practices are eye wateringly horrific. They are environmentally and socially damaging. So we wanted to raise awareness of those sort of issues as well, to encourage better practice. So the compass was developed. As I said, it's a free to access online tool. And it was basically to inform potential end users about the impact of their choices. The other thing that was really, really important at the moment, we don't have anything like the scale of recycling infrastructure that we need to manage all the equipment that's currently in circulation, let alone all the electrical and electronic equipment that will come into the waste stream imminently. We've got a huge problem with this. There's a collection globally and it does depend where you are based globally and as does reuse. Incidentally, people with less expendable income tend to be more frugal by necessity rather than intent, I think or wish. Overall, the collection rate of e-waste is less than 20% globally. And that includes consumer as well as commercial products. But we don't have anything like the infrastructure that we need to recycle this anyway. The recycling processes at the moment, they focus on anything with iron in it. So steel, copper, aluminum, and gold. And there's not very much gold, you know, by mass, it's comparatively little gold in electronic equipment. But of course, its inherent economic value makes it attractive to recycle it. So there are masses and masses of materials, many of which are critical on the critical raw materials list, which aren't recycled. Unless we get our act together and develop a proper recycling infrastructure, there's significant potential for disruption to supply chains.
Gael Duez 38:54
But do you believe it is possible? My understanding of the chipmaking industry, or even slightly less complicated part of the IT industry, the design itself is so complicated, is so… the different metals are melted together to create alloy. You've got ceramic, etcetera, that I don't even know if it's feasible to recycle the way I would say the average John do understands it, which is we will extract to reuse it the same way. And it's more down cycling or it's even. I mean, I honestly wonder if recycling is really the way forward. So don't get me wrong, instead of being able to reuse for super long period of time, as you previously mentioned, having components that are interoperable on open standard etcetera, and saying, okay, you know, this memory card might be 20 years old, so it's ridiculous the amount of data you've got. But hey, I've still got half a billion of them. And if I've put them in some racks, it's still a decent, you know, decent enough, or whatever. But my point is, I think we are fighting an uphill battle if we really want to recycle, like extracting the tantalum you were mentioning, or the cobalt or whatever, rather than redesigning our industrial process and making also sure that the warranty period is so big that so long, that actually we shift the burden of recycling to producer, which are eventually, ultimately responsible for putting things on the market that are absolutely not recyclable. And that would end up being e-waste in a matter of years rather than decades. And when you see the environmental footprint of everything that you describe, we should talk in decades rather than in years. But that's a personal opinion. Sorry, but my point. What do you think about the feasibility of recycling? Or are we talking about a slightly different approach in the recycling industry for the IT equipment?
Deborah Andrews 41:05
I think we need to have all of those things, really. I don't think there's not one size fits all solution. I think there are some really massive challenges with electronic components, for example, because of the way that they function. You're looking at atomic levels, the way that if you're creating signals, etcetera, the way that atoms and the subatomic particles behave. So that is obviously going to limit the way in which components are designed and manufactured at the moment. Maybe in the future they will discover different approaches to data transfer or signaling, switching, whatever it may be. I don't know too much about quantum, and I don't know how this is going to change things, if at all, if we exchange one set of problems for another. I don't know, but I'd certainly like to find out an awful lot more about it.
But coming back to your question about end of life and so forth, I absolutely agree that we need a different approach to the manufacture of many products. I think we need to really focus on the things that can be upgraded, swapped, repaired, etcetera. Again, most or many electronic components, it's nothing. You know, they're so tiny, it's impossible to repair them. So focus on the stuff that we can repair and upgrade and keep in service for as long as possible. The other stuff, we need to certainly develop better recycling capability. But there's an argument that's put forward that some of the materials, the economic value is fine if you have a kilo of stuff, but by mass per component, the mass in individual components and in servers as a whole is very low. So the economic value of any particular materials reclaimed from the server will be low. So we need critical mass to make development of recycling and reclamation technologies for particular materials economically viable. That's going to be the driver. The other big challenge we've got with recycling is sometimes the processes, and we're not. You know, there are lots of ways of doing this. One of the partners in this CEDaCI project, a company called TND or Terra Nova Developments, they developed some new recycling processes to reclaim materials that aren't commonly reclaimed. And they use a mix of thermal and chemical processes. And because you're using more than one process, of course, that increases cost. But, you know, if you have critical mass of stuff coming through the system, it does become economically viable eventually.
Deborah Andrews 44:23
One of the big challenges is very often, but if you have a printed circuit board with a huge number of materials, embodied materials, you're processing to reclaim one or two or three, those processes can damage other materials, and so you can't reclaim those. So, you know, I think at the moment, it's impossible to reclaim all of the materials in PCBs, which are the biggest challenge for recycling. So I think we need to think about when we're designing, not just designing for here now and in use to design, thinking about how can these things be easily disassembled to facilitate, if not, you know, chucking the material into a smelter to recycle it, but actually being able to reuse the materials in the… as soon as we take them off one product, we can put them on another.
Gael Duez 45:22
Do IT equipment manufacturers today start to tackle the issue? Or are they still mostly in the business as usual approach and not at all incorporating, as you said, a repairability aspect when the design sinks, even a bit of a recycling aspect?
Deborah Andrews 45:45
I think there's, and I'm not going to highlight particular companies, but I think there's quite a lot of smoke and mirrors and greenwashing. And this, again, it's anecdotal from personal experience of, you know, I worked through this CEDaCI project. They all still want to sell new equipment, and certain companies say that they have a kind of closed loop, but it's a very open closed loop, shall we say? They're not responsible. They do take equipment, but then they sell it onto secondary market agencies. They're no longer responsible for that equipment once they've sold it on. There's a real need to shift thinking and business models and possibly, rather than selling IT equipment actually to sell services. So, you know, you rent, I don't know whether you do it by compute capability or operating time or whatever. So then they're responsible for maintenance and for what happens, either for extending life, which would be very much to their advantage, or for recycling at the end of life.
Gael Duez 47:00
Which is what is slightly happening with the big three hyperscalers because they've started designing their own servers. And I could bet that they thought about the fact that even from a financial perspective, the longer you keep this equipment, which is on your own cost base, the better it is for your bottom line. And I have a final question for you, because you're privileged witness of the data center industry for almost a decade, two decades almost now, and there is a bit of an elephant in the room that you actually, you teased us several times during the interview about AI and more generally, I would say, about the trends. And I would like to share with you an anecdote while I was recording this very enlightening episode with Professor PS Lee in Singapore, who's one of the best experts in the water cooling techniques and more generally on building energy efficient data centers, especially in tropical climates. We had a really fruitful discussion, and he's a big advocate of technical optimization, and he knows a lot about these topics, etcetera. And at some point in our discussion, he posed, and he had this kind of overwhelmed moment, you know, when you've got just too much weight on your shoulders. He was like, but you know, Gael, at some point we will have to ask also the question of the level of consumption of compute that we want in our society. Because despite all the efforts that I'm doing to reduce energy consumption, the current trends, and especially the AI boom, I don't see how I can make it. And it was this kind of face, like, I simply don't know how I will be able to manage such an exponential consumption in energy, even if overnight all data centers in the world would switch to water cooling, super efficient water cooling techniques and whatnot. And it kind of struck me like even someone within the tech industry and such a strong advocate of technical optimization saying, wow, but the trend is really worrisome. Is it something that you're aligned on, or are you a bit more optimistic about the current trend in the data center industry?
Deborah Andrews 49:29
I think it's really scary. Forget about the ethics and what AI can and can't do. Forget about that. I think it's really scary because it seems that the industry is racing, racing forwards to develop either new data centers, new equipment to manage AI, etcetera. But my feeling is that some of the, what we've learned, good practice, etcetera, is forgotten in that race. So it's a business as usual approach. Build, install, run, replace, run. It's very worrying because the physical resources and energy required to manage AI operate, even the simplest operation, the demand for energy, is astronomical. So there are arguments saying, oh, well, AI is going to measure this, that and the other, and improve, you know, this, that and the other, but, or improve the operational efficiency of this, that and the other. But I don't think we've done any kind of calculations at all to see whether the benefits of running an AI software or, you know, program activity, whatever, to assess resource efficiency, whether the benefits achieved through the resource efficiency or more significant than the impact of running the AI operation. We need to really, and it's not going to happen, but it would be great to pause for a year and just to sort of examine some of these factors and to see where the benefits of AI really lie. So if we do carry on business as usual, and I know there are various regulations coming in from, you know, digital sustainability and so forth in the EU, but I don't think they're going to have a massive impact on, certainly the speed of development of AI is far faster than the implementation of these new regulations. But the other thing that's worrying, if there are constraints of operating in the EU, for example, does it simply mean that providers will go elsewhere, they'll build in locations where the regulations don't apply. And so we move our problems to another part of the world. In effect, you know, more buildings in Africa, Asia, South America.
Gael Duez 52:05
And from a cold financial perspective, do you think that the current trend of building data centers everywhere in increasing compute capacity is sustainable? Or do you foresee some bottleneck or even some forced pose because of resource exhaustion?
Deborah Andrews 52:25
There may be a quantity crunch, unless we think about extending product life, as we've already mentioned, and recycling, reclaiming more materials, changing business practice, business thinking. I think that there's potential for quantity crunch, but whether operators will start to charge for access to digital services. I mean, at the moment, you know, you buy your phone, you have a package, and you can contact anywhere in the world whenever you want. Okay, you pay for apps and so forth, but actually you're not really paying very much for the digital services that enable those apps to function. And a lot of things are free anyway. So will we have to pay for digital services? Is that one way of monitoring or constraining digital activity or not? That's one question. But the other thing is, is it ethical? You know, if we think we're in a luxurious position in Western Europe in terms of economics, although there are people in digital poverty, but, you know, generally, as a child or a university student, if you don't have computing equipment at home, you can go to your study institution and access digital technology there. But that's not the case in many populations in developing countries where digital tech is a luxury for the upper echelons of society. And yet we can see how access to a phone, not smartphone, just an ordinary old fashioned phone, has empowered women, for example, helping them to set up businesses and so forth. Is it right to charge them for data or do we make data charging… is it sort of socially stratified or according to income bracket? I don't know. I think we need to be a lot more visionary, look forwards and be proactive and anticipate problems and design them out before they happen. Now, whether that's… I think that's possible with some equipment, but whether it's possible for, you know, in terms of human behaviour and so forth, there's a whole other matter altogether.
Gael Duez 54:18
It makes total sense. Thanks a lot, Deborah, for joining. That was very enlightening and a unique perspective on the data center industry, environmental footprint and what could be done to reduce it. So thanks a lot. Once again, it was a pleasure to see you and hope to see you to Green IO London as well.
Deborah Andrews 54:36
Thank you very much. It's been an absolute pleasure. Thank you.
Outro 54:42
Thank you for listening to this Green IO episode. If you enjoyed it, please share it via email, on LinkedIn, on Facebook, on Twitter, if you are still there. We are an independent media and word of mouth is the only way to get more listeners. I don't ask you to rate it five stars on Spotify or Apple Podcasts because of course you already did it, didnt you?
It's time for you to grab a book or enjoy a good article. And guess what? You can find many ideas in the latest edition of the Green IO monthly newsletter. And don't forget to book your ticket for the next Green IO conference in London on September 19. Good news for your holiday budget. You can still get a free ticket using the voucher GREENIOVIP. Just make sure to have one before they're all gone. I'm looking forward to meeting you back in six weeks to help you, fellow responsible technologists build a queener digital world.
Roxanne 55:40
One byte at a time.
❤️ Never miss an episode! Hit the subscribe button on the player above and follow us the way you like.
📧 Our Green IO monthly newsletter is also a good way to be notified, as well as getting carefully curated news on digital sustainability packed with exclusive Green IO contents.