 Hi everyone, thanks for joining us and welcome to Open Infra Live, the Open Infrastructure Foundation's weekly, our long interactive show where we share production case studies, open source demos, industry conversations and the latest updates from the global open infrastructure community. We're live here every Thursday, 1400 UTC, streaming live to YouTube, LinkedIn and Facebook. I'm Jimmy MacArthur and I'll be your host for the day to talk about all things green hardware. We're facing hard questions around climate change and a lack of global policy agreement that doesn't seem to rise to this urgent challenge. I'm excited to bring on our speakers today to discuss ways that their organizations are working to change all that. As I mentioned, we're streaming live and we'll be answering questions throughout the show. We'll try to save some time at the end of each episode for Q&A, so please feel free to drop questions in the comments section throughout the show wherever you are watching and we will answer as many as we can. Now let's get going and to kick things off, I'd like to go around and have everyone introduce themselves. Peter, why don't we start with you? Hi everybody, I'm Peter Kulia, work with Ampere Computing and Developer Advocacy and I've been a member of the Open Infrastructure Slash of the Stack Foundation since 2011. Excellent. Laurent? Hi everybody, I'm Laurent Mademazar. I'm the CTO of the Co-Founder of Exion, which is a new EDF group subsidiary, a big French utility company for those who don't know. It's rep-riding cloud services around blockchain technology and intensive workloads, computing such as HPC and AI. Excellent. Zach, go ahead. Thanks Jimmy. My name is Zachary Smith. I work at Equinix, I run the Equinix Metal Business Unit, where we work to make it easier for our customers to deploy physical infrastructure across our 200-plus data centers and 64 markets around the world and turn it off and on easier. Very interested in the sustainability conversations here today. Excellent. And last but not least, John, why don't you go ahead. Hi everybody, John Miranda. I work at Intel Corporation. I work in Intel's Data Center and AI Strategy Office, where we track a number of topics, but sustainability has risen to the forefront. It's one of the topics we're also following, so I'm happy to be here. Excellent. Well, thanks everybody. I'm happy to have you here today as well. John, we'll go ahead and start with you. Just kind of high level, talk to us about the green computing movement and how open source initiatives like OCP and putting into this. Yeah, it's on the movement part. I mean, we've been, you know, in general, society's been talking about sustainability for decades. It's not particularly a new topic, but what we really noticed is a real sharp accent in the focus this topic is receiving, even in the last year or so. So if you look at, you know, there's rising regulatory pressure in the industry. There's investor sentiment. Investor pressure is on the rise. So the expectations for the industry to do more are rising as well. So in terms of the green movement, you know, we've seen CEOs of a large, large corporations are all declaring goals such as carbon neutrality, but then the question arises, well, how do you go about doing that? And I think that's the phase we're entering where people are starting to unpack, okay, that the challenges are being issued, but in practice, then, you know, it's a broad topic and how do you start decomposing it? What actions do you take? And I think what many of us are realizing is no one company, we can all do our part, but no one company can turn this ecosystem in of its own. So, you know, so when you start looking at movements like OCP, it allows us to come together as a community and drive more change. Speaking about OCP, so on November 9th and 10th is their Global Summit and they're going to have a whole track, a whole day dedicated to sustainability. So it includes, you know, information sharing and kind of engaging a community to drive more action in that space. Yeah, absolutely. And, you know, you mentioned how do we do it? We tried to get a wide variety of folks here today. You know, we've got a couple of chip makers in Imperial and Intel. We've got Equinex on the data center side and Exion, who's got an interesting story around recycling HPC computers and giving them a second life. You know, so talk to me, you know, Peter, you're a member of both OCP and GCP and, you know, Ampere is working to build dense data centers at the edge with more efficient chips. Like, what are the impacts that you're seeing in the work that you're doing and is it making a difference? Yeah, I would definitely say the stuff that Ampere is doing is making a difference. And I can talk about it from, you know, hands-on experience from what, you know, operators see in the data center once they are utilizing, you know, Ampere's chips in practice in reality, right? So, you know, and to just give a little talk about that, I work directly with the Oregon State University Open Source Labs. The OpenStack project has a leg of the CI, for example, when they're on our first-generation processor. When I was working with the operator there, Lance, to stand up the infrastructure that we have there when the power was first applied. And I have to preface this by saying, you know, they also host infrastructure for the open power communities with an open source as well as other, you know, other infrastructure for bills and things like that. And that has a good perception of the power requirements down to the rack level for different platform architectures, right? And when he was there and lit up our racks, the rack of 12 of our first-gen servers the first time, essentially they used less electricity than the top of rack switches, right? And because of that, he thought something was broken and he had to go get someone else to confirm because it's not, you know, that's not typical. So from, you know, what it means in reality, right? Like, as the infrastructure is changing, the demands are changing, they're requiring more power, you know, as we move forward, the physical space that this equipment is going to consume isn't changing, right? The power going into that physical space isn't necessarily changing, especially, you know, when some of these data centers are on the edge for telco, right? Some of these buildings have been around for a long period of time. They have physical limitations in terms of, you know, what is a, what the available resources that could be pulled in. So, you know, efficiencies in maximizing that efficiency for, you know, the current model will help in both, you know, reducing the, not just the electrical, you know, consumption that we have today, but it essentially puts us on a path going forward to try to, you know, really optimize that the data center stamp to take advantage of what's in there today and help us plan to go forward. So, yeah, that's probably one-winded, but a little bit of everything. Yeah, that's great. I mean, and, you know, speaking of size limitations, I think that, you know, Zach, this is something you're tackling, head on with the Open 19 initiative. You know, it's, you mentioned to me, it's more than just hardware, it's the rack size, it's the packaging, it's everything, everything around the hardware, too, that's important for these great initiatives. So, tell us how Equimax fits into all this and what you're doing at the Open 19 initiative. Yeah, thank you. I think the, you know, both Peter and John point on the complexity of this challenge, right, which is that this is an ecosystem that, you know, frankly, I think I want to pat ourselves on the back and say, you know, from the incredible technology that companies like Ampere and Intel create to how we get it into market, to how we enable it to be used by customers worldwide. We've made it pretty good, but it's actually an incredibly complex industry with lots of parts of the value chain that have to kind of evolve to help customers towards a more sustainable use of, you know, our capabilities and our technology in general. And one of the things which Equinix is the world's largest interconnected data center operator, we have a goal here at Equinix, a stated target of carbon neutral by 2030 using science targets. We're seeing our customers continue to ask us for more help. They're pushing there to no longer somewhere deep in the RFP that there should be some like ESG, you know, side benefits or marketing. It's like number one on top, you know, requests or demands from all of our Fortune 500 and Fortune 5000 customers saying, we need more help towards reducing our carbon impact, making our technology more sustainable, using it longer, all sorts of other things that become part of that circular thing. And so what we're focused on at Equinix is, you know, we see a lot of the technology land at Equinix. Unfortunately, we aren't in a hyperscale model where we can control what type of equipment our customers want to put in our data centers. And so we've sought out ways to work in open communities like Open 19 as part of Linux Foundation where we can help open standards around form factor, right? And so what we're thinking about here is primarily the advancements we need to make in conjunction with silicon and server manufacturers that also ripple into the broader retail data center. So think of things like more efficient use of power like what Peter was speaking about, chips are getting hotter, we're having to put more power in smaller footprints. How are we going to do that more efficiently? So having more efficient use of power, power supplies, conversion, you know, different kinds of software to find power that we can put the right type of power in the right place at the right time. And then cooling. And cooling is a huge portion of our world, the waste that we have or the impact that we have in use of water with evaporative cooling. We operate all around the world, some places that you can leverage outside air, some that you cannot at all. And we're getting into situations where we have a significant amount of heat that we have to remove from our data centers. So the question is now how can we use that? So looking at a standardized liquid cooling coupler for plate cooling or single or two phase cooling, how can we do that with facility scale, you know, in the large scale facility footprints we deploy that can work with a variety of server and silicon manufacturers so that we can adopt much more quickly an efficient way of using liquid cooling not only to remove heat from the server, reduce things like fan waste, improve performance, you know, on the silicon that is there, reduce evaporative cooling or evaporative water use where we can actually have a closed loop system. And then frankly one of our goals is, well I think it's not our goal to requirement in order to hit our carbon neutral science targets is we have to capture that heat. We have to use it to make new energy, sell it back to other places. It is currently a very, you know, big waste that we do mainly because we can't capture it. And so there's a lot of work that we have to do with our suppliers and a broad ecosystem. No one person can do it on their own. It's not possible, at least for the thousands of enterprises that work in, you know, I'm going to call it the heterogeneity of multi-vendor, multi-location, you know, footprints. So what we've done with open 19 is just work towards a standard on form factor within the rack, mainly thinking about how do we have a liquid cooling coupler that we can scale across our footprint that other, you know, OEM and server manufacturers can support and how do we move towards more efficient use of power? Yeah, I love that. I mean, as everything is moving to the cloud, everything is practically moved to the cloud already and it's growing and growing, you know, having these efficiencies are going to be so important for, you know, every country in the world. There's no other way. We got to make it better. In France, they have where Exion is located. Laurent was telling me that there's a, forgive me for my French, but a raison d'etre which is mandatory for every organization in France. So Laurent, tell us about Exion and your reason for existing. This is the group, the idea of group raison d'etre. This is to build a net zero energy future with electricity and innovative solutions and services to help save the planet and drive well-being and economic development. This is a big challenge. But we have to decline this in our domain. So for us, firstly, we use as many as possible low-carbon electricity produced by our parent company. We also, as you say, give a second life to existing hardware because around 75% of carbon footprint of IT equipment is due to producing and to manage the end-of-life of equipment. And at EDF, we use big HPC systems, which was in top 500 classmates. So thousands of x86 servers, high throughput and low latency networks, and multiple petabytes of storage array. So we decide to rebuild a solution with this three or four-year-old hardware. But we need to boost the performance. And for this, we add chips and accelerators to existing hardware. We also work on technologies such as immersion cooling to address the heat problem. We also work with the EDF group subsidiaries, which is named Dalkia, around how to reuse the heat to heat pools or farming usages or other kinds of needs of thermal calorites. We also provide for services information around CO2 impacts. So that's like a fund bill where you can have all the details of I use this service and I have the carbon footprint impact to sensibilize the users around their own impact every time there's not the same carbon footprint impact. There's a moment where the electricity mix is made by carbon power plants. And we can give this information to our clients to say, is it the right moment to launch a big HPC workload or can you delay for one, two, three or four hours when the mix will be better? Yes. I love this idea first of making your customers aware of what's going on and what their individual impact is and also addressing things like time of day, time that you're using things. I can't remember who was talking about it, but someone was discussing the fact that ideally your refrigerator would only make ice first thing in the morning or overnight, right? And during the day, it would turn off and you wouldn't make any other ice. These little types of tweaks can be done through technology. And it's awesome to see this. We talked also about the circular economy and I think Exion fits in that. Zach, what are your thoughts on, you know, right to repair and the way I see it, you kind of leading the way in legislation that, you know, that is helping climate change and helping address those things as opposed to the U.S. where you're sort of lacking that. Yeah, I think Laurent put on a few really interesting things and I drew myself a little circle here because there's all these components and the biggest one that you focus on for me is like, how can we have less of a, you know, a carbon footprint using technology? And the answer is not make the technology as often, use it longer, right? Use it for an extended lifetime. And I think there's two really interesting opportunities there. One, which is kind of in our control. I'm going to say easier and that's to, at least like within an Equinix data center, for example, we're looking at ways at how we can take the rack level infrastructure of cabling and power and sheet metal and have it last much, much longer versus tearing it out every time we want to upgrade the computer, right? And that's what we currently do today is with our kind of typical X86 pizza boxes, you tear out the power supplies and the, you know, the rack level, you know, networking and, you know, cabling and related every time you want to upgrade the computer, which is on a very, very fast refresh cycle in many cases while the physical infrastructure side is not on a fast refresh cycle. So having that extension of the physical rack level infrastructure is a big goal of what we can do. And then this other one is actually a business model of Silicon and OEMs, which currently right now is really incentivized towards helping you buy newer technology on a regular basis. And I think one of the big challenges we have as a community is move this, what you were talking about circular economy. It's going to be a requirement for people like John at Intel and Peter and I'm here to have business models as well as our OEMs, which get paid, you know, when the technology is used longer. And right now that's not currently the economic incentive. And so I think moving that alignment is going to be a critical part to make sure that this technology has its full lifecycle and that there is an incentive for it to come back in the right way to be reused, recycled, put to another customer use case, etc. Which today is not merely as evident. Yeah, excellent. And Peter and John, he called you out specifically. So what are your thoughts on that? As far as I think some of these things to the services and just generally adjusting the business models of Silicon makers? Yeah, I had a network hiccup, so I'm back in. Yeah, so we talked about efficiency and I heard the tail end of circularity, but I think that's where the hockey puck is headed in terms of the conversation moving from that linear model where we extract, we make something, we use it and we discard it to circular. And circular starts at design and it's both hardware and software. So we need to think about the service life, the upgrade ability. When a software support stops, does the thing become a brick? Is it modular? Is it multi-generational, upgradable? How long can that bomb go on? Can it be enabled into a secondary or tertiary life? So all these concepts begin at design. And so I think as an industry, we need to move towards that direction. Peter, thoughts? Yeah, and I guess to add to that, Ampere has been committed to disrupting our slice of the industry since our inception. The way we architect our chips, the way that we think about our platform lines and those sorts of things are all trying to remove the legacy kind of way of thinking about it and actually be disruptive in this space. So along the lines of some of the stuff that John just mentioned, we ourselves have today an RMSOC that is capable of having a socket, removing one chip from the socket and being upgraded to another, right? Which isn't typical for those types of things. So we are thinking about a lot of the points that John mentioned, as well as thinking about ways we can achieve efficiencies or disrupt the industry through some other mechanism that will directly affect the performance of the system, the efficiency of the system, and the density of those cores in the rack themselves to help carve a path for the future of this. But as it was mentioned by everybody here, there's a million levers to flip across lots of different things. I can definitely say Ampere is committed to doing our part and it's been a fundamental pillar of our existence from our inception. So yeah, I look forward to seeing and working with the rest of these guys. How do you think? Peter, you triggered a thought that the million levers to pull. So I think one third angle that's really interesting that our industry can make some headway in is the idea that as we add renewable energy to the grid, it becomes harder and harder to absorb it because of its variable nature when the sun shines or the wind blows. And yet the grid has to maintain stability between supply and demand all day long. So what role can technology play to behave as a virtual battery so you can, is there more work you can pre-do or work you can defer and align that workload intensity to when it's sunniest or windiest and then relax that energy consumption at 5 p.m. when your energy profile is more fossil-based. And that can extend to every device in our lives from down to the refrigerator that can make extra ice cubes at 10 in the morning and then the compressor works less at 5 p.m. And as far as the grid's concerned, your refrigerator is acting like a virtual battery. So as we think about, you know, all the computation that's happening across all our devices, I think there's a transformation happening in energy as it's becoming more distributed. We harvest our energy, we store energy, we use our energy. So all these nano-grids and what role can compute play to exploit and bring this energy more rapidly to the grid because if we're using solar energy, it's okay. What we're trying to do is avoid using fossil energy. So what role can we play to kind of accelerate that movement I think is an interesting aspect to this whole challenge? Yeah, absolutely. I love that concept, especially as we're moving more towards IoT and Edge, you know, using every device as a battery is kind of an incredible concept. Jimmy, maybe I can add a point there because Laurent touched on it about, like, exposing information, right? We don't do a great job in our industry about even having standardized ways or any kind of way in many cases to tell our end users who might be thinking of deploying the proverbial ice-making job, e.g. running their, you know, machine learning training job. We don't have a great way of telling them. Like, now's not a great time but now's a more carbon-heavy time or whatnot, or this region might have more capacity for you to do that in a better way. And so I think one of the areas of opportunities we have is really around figuring out standardized ways versus, like, you know, individual point solutions around exposing, you know, the state of the energy and the carbon footprint of the resources you might be able to consume and allowing you to do something about that. As people become more sophisticated about workload management with their software paradigms, you know, orchestration, other things, how do we upstream, you know, the developer experiences that are going to basically control this from this, you know, pretty complex set of inputs behind the scenes. My comment to that is welcome to the matrix, right? The reality of what you're talking about is kind of the direction that it's all going in, right? With the whole movement within IoT and collecting the information we need to be able to compute on it to make intelligent decisions that we then hit APIs that, you know, trigger some level of automation to, you know, move the chess pieces around the board to achieve those goals, right? Like, a lot of the fundamental Legos of each shade of color being built, you know, kind of as we seek towards that model, right? But as your point being, until we have, you know, the right technology in place to bring the information to the forefront, we're not even necessarily sure what we're dealing with. And although some of the other pieces in the background are there, you know, we have to still do all the work to couple them together to get to a closed loop, right? To achieve the level of automation to get to the, to realize that dream, right? So yeah, I look at that as like, when we're there it's almost like it's well, we'll be in the matrix, I assume because it'll be managing itself. So, you know, Zach mentioned Laurent and it raised the point to me that I mentioned earlier EU legislation, in my opinion is miles ahead of US legislation and regulation on climate change and climate policy. So I think it's fair to say that EU is leading the way. And as a result, at least with OpenSAC, we've seen a huge explosion in public cloud growth because of data sovereignty and rules and regulations in that that area. But Laurent, I'm curious, you know, the carbon reporting that we're doing, do you see something like that becoming a part of EU regulation moving forward? Oh, the this is a premise of the of the regulation today that's not mandatory, but I think tomorrow it will be I've heard that will be next year, but I'm not sure that every provider called the needs but this is the I think this is the future and we need to do that because we don't have the choice and also we need technology to do that. We talk about the idea of the matrix, but we need to have machine-to-machine protocol a full, optimized complex system and for this we believe that blockchain technology could be a part of the solution to track and to automatize the contract between all the participants and to incentive the end user or the company because this is a we need to deal with the economic system we need to deal with so we have to make an automatized system which could address that. I've just changed the sense of the question but I know that regulation is coming in Europe and for our US based companies, ampere and equinex and intel you're obviously global as well but what do you think about the EU legislation? Is it something that you're working hard at your organizations to change at a manufacturing level and at a business level towards these green initiatives but should our businesses be driving legislation and lobbying Congress to make those changes in the US? As a global company it will affect large US multinationals and I think has been stated if you look at the energy footprint of computer and data centers by the end of the decade the estimates are very widely but by some estimates it could be 20% of global energy will be consumed by the ICT sector so it really behooves us on us to start bending that curve so to speak as we're seeing nascent actions in the EU but we're also seeing actions in the United States some states now for some gaming rigs when you go to order them from a major OEM if you're in one of 10 states you can't order it because it doesn't meet efficiency standards so it's affecting the PC business already in the United States so it's not just the EU thing so I think that we're going to see more of this so as an industry I think it's important for us to take the lead versus being in a reactionary mode yeah I totally agree and we're seeing it all across our footprint we're also very global as an operator and we take it very seriously hence why we signed ourselves up we're the first major data center company to use renewable power but exclusively throughout our portfolio but it's still not enough so we set our science targets for 2030 knowing we want to help lead that that's our responsibility to our world and our customers I think one area I wanted to react on really quickly was maybe ponder a question for the group which is we have struggled sometimes we're seeing it right now where energy prices due to this transition to renewables in Europe is very variable so we're having these internal debates around what to do about that do we eat it on behalf of our customers are we just do we help expose it in ways and give options to react what do we do and I think a question that we have is we see that potentially the kind of sustainability and technology thing is moving past table stakes to do business with us we've always seen it as a branding kind of part of our brand which is it's on brand for us to care about a sustainable world and so we invested that but our customers often expect it for free and what we're kind of getting into is that it's not going to be free like as we move to net zero and to negative it's not free there's actually a premium for a lot of these changes have you seen any impact from customers and how are you dealing with that do you think customers in the end are willing to pay for a more sustainable future or do they just want it to happen with supply economics so everything else around that's a great question one area I was thinking about is for some of our customers they're also trying to meet their carbon neutrality goals and one way they can do that is to buy offsets and there's different qualities of offsets so there's a cost to them to achieve their goal and it's inclusive of their IT footprint so if you offer them a more and by the way as more of us pursue these instruments the demand is going to outstrip the supply so the cost of buying these offsets is going to rise because there's more of us trying to buy the same thing so they're going to incur a cost no matter what so the question is is there a role that the equinexes and hyperskillers can play to hey we can offer you a more attractive way to meet your carbon neutrality goals that may be more TCO friendly than you trying to solve your own problem yourself directly and I think to comment on that I think some of them are that's why I think Ampere has a lot of opportunity and there's other factors that play too like the last two years definitely helped the rise of more consumption in the cloud is making them rethink literally how they build their data centers because now everything has to be more efficient as the majority of us on this panel are in the US we're built on 100 plus year old infrastructure there's tons of efficiency that could be done along every level of delivering power from the existing especially for our folks who may be in Texas from the existing power infrastructure and grids into the data center so if we look at all the technological enhancements that can be made to make the flow of power into the data center more efficient the creation of that power more efficient and ultimately the computing using that power more efficient I think it's going to force it well it's already partially forcing the issue both in regulation in government through just current discussions that are being had as well as what's being teed up for the the next wave and it's all from my perspective accumulating to a head you couple this with the current security industry issues and I think we are on the precipice of a wave of massive change in the data center that's being driven from those multiple points and coming to a convergence because much like you know securing things like you said it's not cheap to make things efficient it's not cheap to secure it either so we're going to get to a point where there will be a penalty I believe if you're not going to do that stuff because there's inherent risk to the cloud operators or service providers from their customers who aren't taking a proactive approach and that increases the liability from my perspective standardization is key regulation is going to come out of either it's going to help drive the standardization or in some cases we're seeing the industry driving its own standardization because the customers driving their own standardization like in the case of OCP to force the industry to make change think about it OCP has been around for quite some time now how many in the last few years are we seeing the effects trickle into consumer level devices where you can get OCP compliant hardware that is off the shelf right so it obviously is going to take a long time I have a comment on that which is I think especially for the broader enterprises versus the pure digital service providers we have to figure out how to involve OEMs absolutely that's where the OEM the OCP construct is really going to be difficult to bring OEMs along OEMs yes but OEMs it's really hard because it's the IP ownership structure really is meant to remove a lot of their value chain and so I think just given how we see technology reach enterprises we need to figure out how to bring OEMs and other kind of what I'm going to call proprietary technology companies who monetize their technology in that way along for the ride welcome to disruption right that's why we're all you know working in open source because we've all agreed to standardize on some level of technology and move ahead on where we can compete right so it's going to happen and I think in the rest you know that's going to happen in other parts of the industry as well there's going to have to be a standardization for evolution to occur right and we're all going to have to give something up to get to that so you know when we all agree here that it's you know at some level it's necessary and it's going to be and at others it's going to be mandatory so you know it was it was touched on earlier on standardization maybe one area that we can all agree on is the importance of measurement because if you can't measure it you really can't manage it and if I you know like let's say I'm a medium-sized business and I run payroll for thousand employees how much energy what was my carbon footprint like and would it have mattered if I run it Tuesday at 7 a.m. or Tuesday at 4 p.m. or would it have mattered if I ran it in the data center in Atlanta versus a data center in another part of the country so how do we get to a point where we can expose to those end compute users the carbon footprint and motivate them maybe it's a form of congestion pricey fuel motivate them to to make the right decisions they may want to make the right decisions but without the measurement and the data they're in the dark and so I think I totally agree John and I you know I've been inspired by like what stripes done with their climate project where they have made it so consumer simple for businesses to turn on one percent of their transactions to go toward climate you know carbon capture which is you know not a very at scale market but they've made it so productized and that's something where I think we have a lot of work to do is taking that information and productizing it as part of what we do and saying here's how you can use it right in a more repeatable manner versus maybe some of us who are highly passionate and involved in doing this as our job right there's a lot of enterprises like the payroll you know running of the payroll software which they just need it productized so that they can use it that information so we've got a quick question coming in from from the audience Thomas Coy asks the raid boxes we aim to be CO2 neutral the idea of processing data where the footprint is better sounds interesting but it's a very complex gap to solve in a hosting setup yes anybody want to take that one Peter Zach you know I'll comment you know that's part of the opportunity for you know companies out there right is solving those types of problems yes as we if we want to be able to you know move compute workload at ease you know dynamically on demand we not only need the data but we need the technology to support it right you know we've talked about things like hub bursting for like 10 years yet but still the seamlessness and what that really means isn't isn't fully realized yet right so from a mechanism perspective in order to enable you know kind of those things the technology has to evolve and it has to present itself and there has to be companies and providers working together to ensure that the you know the doors and are open and capable of handling those types of you know activities right so there's a big technological aspect that has to be enabled to get to the point where you can you know shift in dynamically move you know workloads globally as you know to meet to meet your pricing and that's where the opportunity is going to be lies for the industry to you know capitalize on Laurent any thoughts on that oh yes I think it's could be maybe done by decentralizing the IT system we can do this I know that we can do this now with bigger cloud providers services and we need to be deployed on different sites and different maybe countries different grids electrical grids to do this technically I think it should necessary should be done on an open system because we need to to interpret to have the data to interpret and to be ready to start I don't think that the workload should just move but it could be build and launched on the in the good place I have a piece of feedback on that which is that you know we use a term internally at Equinix called how to move physical infrastructure at software speed when in reality these are very very physical components right building data centers takes time ordering computers is a long lead time item installing that is very complicated and what we do right now is an industry is we really work around a kind of like how do we get things into the data center and a lot of the innovation has been done in hyperscale which is how do you get a lot of the same thing into a few data centers and we have this other problem which is how do we get a few things in the many places or how do we move them more efficiently and cheaply from one physical location to another like the whole concept of rack and roll or you know taking your entire pre-built you know rack and putting it in and it works for a very small subset of the world and then the efficiencies we see it in Equinix because we charge people hundreds of dollars an hour to kind of move a cable right or to put a server into a rack it's extremely expensive and it's not standardized at all there's no standardized cabling understanding or power and as we get more complex with the power of the cooling and the related we have an even harder problem I look at it as the difference between making completely craft coffee versus an espresso capsule right we don't have any sort of espresso capsule for data centers because everything is bespoke once it gets into a 19 inch rack and so I think there's some big opportunity to address Thomas' challenge of right now once you put equipment into one place it is almost impossible or not at all worth it to move it to another place and we have a lot of work to do where we can make that technology infinitely more reusable but we don't need it as much there let's move it someplace else oh this customer doesn't need it anymore let's reuse it in another place or hey we put it in and it's time to refurbish it and add something to it let's make that more circular right now so much of the friction is actually in the installation because we've never thought about how we get it out and reuse it in a kind of reasonable manner I'm having pets versus cattle flashback right now because that's essentially what we're talking about right disaggregating everything from the hardware itself right to provide a layer of flexibility and movement yeah I think there's a technology approach to it which is I believe you and Intel and related are working on awesome ways to re-envision our architecture to be more fungible throughout things and then we also have a physical thing which is just how do we make form factor more portable and less friction free and this can get itself down to in our industry like a standardized reusable packaging model which we don't have right now we have a lot of cardboard that goes into our data centers which we then throw immediately out it is impossible to repackage those computers and put them anywhere else it's very very expensive and so even just getting into like working on standardized packaging could remove a lot of this friction that currently doesn't allow us to reuse components I'd say the software side this is where OpenSec comes in is we're trying to solve this through interoperability and being able to lift and shift your entire workload over to another set of VMs another set of servers ideally where I think where we struggle is still at the public cloud level not being able to move from one public cloud to another because of federation issues and metering and monitoring issues that I think still need to be solved but in an ideal world everyone is running OpenStack and we can just run anybody's workload on any public cloud or private cloud that happens to also be running OpenStack and then we kind of solve some of that hardware problem as well I want to make a comment to one of Zach's points earlier you mentioned about adoption of a lot of the kind of advancements that were made from the hyperscalers that's driving some of the technology along those lines I think one of the things we need to do much like OpenStack when it came about to address the idea of scale of just another scale problem advancements are being made by hyperscalers we need to scale them down to be applicable to enterprises edge other use cases I guess cherry pick the things that work best and figure out ways to apply them to the smaller stamps that are out there and I think once that's going to as things typically happen that will eventually happen but it always takes time in cycles that's a great point this is the solar sales reach point where now it can be widely addressable by my neighbor to put on their roof these has to happen at some point in these larger more complex systems as well for technology I know I was going to say the interoperability is going to be really important we talk always about the public cloud but the reality is the end result for enterprises is going to be hybrid there are some things that will not go in a public cloud you got major manufacturing sites the data is there you need to maintain that kind of isolation of keeping the compute localized to where the manufacturing process is happening so we got to think in terms of a hybrid end state which is inclusive of multi-cloud interoperability is going to play an important role there and that can also be connected back to sustainability I think probably time for one last question here John from Texas is coming on with how do the opposing forces of standardized hardware and rack configurations versus increased demand for specialized hardware etc play into that push for portability I would love to take that one for 30 seconds I personally believe we're in a more although it would be convenient for us to say that we're going to have a homogenous infrastructure you know substrate that we could really build this scale around and whatnot that we're actually going to have a very heterogeneous infrastructure world that part of our job here is like as workload becomes more important we're going to Intel and ampere are going to make specific chips that do one thing really well or do one thing best right like we're going to have accelerators and unique workload that happens much more I'm going to call it not bell curve like right there's going to be the anti part of the bell curve and I think that that's a huge opportunity we have to recognize it versus push against it because right now in compute we have this kind of commoditized 80% should run on all the same stuff thing and it actually might flip where more stuff Laurent is looking like HPC and less of it is looking generic and so I believe that we have to work that into our standardization not in terms of all computers look like this or all computer racks look like this or whatnot and hence why I'm very focused on form factor and I take inspiration from the PC world of ATX cases you could put almost anything in an ATX form factor and everybody knew how to build an ecosystem around that if we could build an ATX case for you know the mechanical sides of racks and whatnot then we wouldn't be flipping and moving our power supplies from left to right and flipping and moving our you know cables on our Nick from here to there which complex you know creates massive complexity in the in rack data center side so maybe it's more think about the area that we standardize in is it in the ATX case of the rack or wherever and then allow lots of diversity within that I think probably ampere and Intel can speak more to being able to more fungibly use disaggregated components across the network or whatnot that would probably make this a lot easier yeah I think sorry automation the things like plugging how easy it is to plug in a cable when you're automating entire data centers that increases your failure rate of precision your chances of screwing up pulling the MAC address off that all those things increase consistency and precision around those things in the data center definitely affect your ability to operate it at a high level and will affect the cost of that through the suck of time in effort to go and troubleshoot and find those issues because of the lack of standardization I think sorry I wanted to add to Zach's point on the case thinking about modular designs which are more kind of you can upgrade a piece but leave the rest of it alone because there's a carbon footprint and the fans of assembly the power delivery all the stuff that goes around the actual FPGA CPU can you preserve all that even as you upgrade or modify one of those components so I think thinking about more modular designs is an important aspect here well this has been an amazing discussion I think we're about out of time but I want to thank you all for coming on a really interesting and super thoughtful discussion so with that we have another great episode coming up next Thursday focused on the hybrid cloud kicking off a series of episodes to talk about organizations deploying hybrid cloud environments and the technologies they're working with and the vendors who are supporting the production use cases and we've also got a very special two day episode in November and we hope you can join us for open in for live keynotes on November 17th and November 18th our biggest event of the year so register now at openinford.dev slash live slash keynotes and I want to say again thanks to our guests I appreciate Zach, John Laurent and Peter for coming on and it's been a real pleasure thank you and remember if you have an idea for a future episode we want to hear from you submit your suggestions to ideas.openinfo.live and maybe we'll see you on a future show mark your calendars and I hope you're able to join us next Thursday at 1400 UTC thanks again and have a great afternoon