Guy Daniels, TelecomTV (00:23):
Hello and welcome back to the final two sessions of this year's DSP Leaders World Forum. And we start with next generation infrastructure and scaling networks for innovation and business agility. A greater reliance on data centers and an increasing need for AI compute are driving changes to network infra coupled with an ongoing requirement to improve business agility and service scalability. They also need to be highly resilient and of course secure. So what are the best strategies for ensuring next generation digital infrastructure? And of course we need next generation digital infrastructure for our next session on six G. Yeah, without the infrastructure whether does that leave six G? So very important session. This one I'm going to ask our guests to introduce themselves starting on my far left with Matt.

Matt Rees, Neos Networks (01:23):
Hi. So Matt Rees, do a combined CTO/COOrole for Neos Networks. If there's anyone in the room who doesn't know Neos, so we UK B2B connectivity provider, so fixed line connectivity provider. So hopefully that helps for anyone who didn't know us.

Guy Daniels, TelecomTV (01:41):
Thanks Matt.

Eric Van Vliet, Dell Technologies (01:42):
Sure. Eric Van Vliet, Dell Technologies. I work for the telecom systems business group that I believe Manish introduced yesterday. I'm responsible for strategy and market development in emea and I work with major telcos, other telcos that are not always seen as major telcos, but also how can we bridge the gap between the enterprise requirements and the telco network and the infrastructure. And obviously we're an infrastructure company, so this is a great session for us.

Guy Daniels, TelecomTV (02:12):
Excellent. Thanks Eric.

Yue Wang, China Telecom (02:14):
Yue Wang chief technologist of network and AI in China Telecom.

Guy Daniels, TelecomTV (02:19):
Thanks Yue. Thanks very much indeed. Thanks Oliver. Be good to meet you all. Let's start with our first talking point of this session and we want to try and understand what's driving investment in next generation. What external demands will dictate the bulk of telco infrastructure CapEx over the next five year period? How should operators deal with this? What are the demands? Matt, I'd love to start with you and get your perspective from the UK network perspective, but what external demands are you seeing that's driving evolution?

Matt Rees, Neos Networks (02:57):
Well, I guess I can answer both from my own perspective and I think everyone who's been on this panel or these panels for the last two days has done a great job of introing that for me, haven't they? So I guess everything we've heard about 5G earlier on in the sessions today and rollout of new ran and densification, less relevant to me that one. So these are no particular order fiber, clearly it's the foundational underpinning infrastructure for pretty much everything that we do for our customer base. I guess more so now in the core as well as I guess over the last five year period, I guess all the focus has been around access FTTP deployments, isn't it? And I think we'll still see bits there, hopefully a little bit less duplicated overbuild, but I think clearly core now becoming more and more relevant with the need to connect high speed, low latency data center connectivity.

(03:47):
Five is still the fundamental building block for that heard about ai. I guess that links to the data center point. I guess that's then both the investments that we'll need to make in enabling and in software for us to consume and make our own businesses more efficient. And then I guess you can't not talk about the need to continue to grow, build on our sustainability and security guess clearly as the world evolves, security remains a persistent threat and you have to continue to allocate a chunk of your CapEx spend to making sure that the network remains safe and secure.

Guy Daniels, TelecomTV (04:24):
Absolutely. Thanks very much Matt. Eric, from your perspective as a supplier through the industry, what demands are you seeing?

Eric Van Vliet, Dell Technologies (04:33):
Well, I think if you're looking at the networks that telcos have built, they're very much been about consuming of data for everyone. And we've seen the trend of upstream becoming more important. I think somebody said upstream is becoming the new currency earlier, early today. I think it's going to be more and more about the experience and what people are going to do with the network than actually having a big pipe because out of my own experience, but I think everyone has good enough connectivity the majority of the time. So I think the majority of the CapEx investment other than keeping the network running, like you said, met, it's going to be the use cases that generate the data. So it could be use cases around that's machine learning, artificial intelligence experience, anything that the subscriber or the enterprise creates and does not want to rely on an hyperscaler.

(05:34):
And we see edge becoming more and more important because even though it's not a low latency application, latency within, when you do things like generative AI or even when you're looking at IOT applications, latency is becoming more and more important because we're making more quicker decisions based on the consuming, sorry, based on the generated AI or other answers that we get through our models. So I think yeah, the CapEx investment will be all about, okay, well what are my customers doing with my network? And that's why I'm investing. It's no longer the need, the fiber wallets are getting there, everything's sort of done and 5G is now mature is a big word, but it is there. It's been there for a while. And again, the network's good network's, strong, reliable. So I think the investment's really about the use cases, a lot of augmented reality. I think we'll see happening. A lot of devising wearing wearables. I think that's really what it will be about.

Guy Daniels, TelecomTV (06:36):
Great. And Yue, do you also see the external demands as being around use cases and experience rather than top line speeds?

Yue Wang, China Telecom (06:47):
Yeah, I think for me, I think the infrastructure advancement is one to support the ai and secondly is to build the differentiated services. I mean I really agree with what Eric just said. Traditionally, teleco has evaluated their business success by the volume of data, more users, more devices, more data, more traffic. But we do see with the integration of ai, we do see there are more opportunities to actually provide the differentiated services to the third party. And I think that's a big opportunity for the operator. But having said that, in order to provide that better quality of experience to the user and to provide that differentiated service, you really need to have that infrastructure that is able to support AI really nicely. I think in the previous panel yesterday there was a discussion about how AI is going to change the network traffic. I think it's probably just not only about the increase of the network traffic potentially generated by ai, but it's probably more about the type, the behavior of the traffic that is going to be changed by ai. So one of the example is for example, the data center backhoe. So what we have observed is with the AI traffic, they have changed the behavior of the traffic from North source traffic to the west to east and data center backhaul is something we have tried, explored, and also generated this what we call the super fast compute line. So the idea behind this is you can essentially change the network to become solid compute fabric and this is where that AI infrastructure comes from.

Guy Daniels, TelecomTV (09:02):
Great, thanks very much. And Matt, let me just bring you back in on this one. And just picking up on what's being said with your customers on your network, do you also see this move away from a constant speed evolution to experiences and use cases as we've been talking about here?

Matt Rees, Neos Networks (09:20):
I guess our products largely connectivity a bit commoditized I guess. So really we have to try and differentiate on experience. A lot of the ethernet market in which we plan in the uk, the price delta is relatively small. So I think yes, we absolutely have to give customers that simple transaction. How do we make, in my case, nios networks the easiest buying option and the easiest in those rare occasions where customers might need help. Again, make that a simple dialogue I guess. Absolutely. The way to stand out in a crowded market is differentiate experience, right?

Guy Daniels, TelecomTV (09:59):
Yeah, absolutely. Thanks very much Matt and Yue, lemme come back to you of what you were saying about how AI is shaping traffic on the network and shaping the infrastructure. Because you look at reports and there's huge numbers being thrown around and what the opportunity is for telcos with AI traffic for AI infrastructure. Is there a direct correlation between AI traffic and for telcos to support AI infrastructure directly?

Yue Wang, China Telecom (10:28):
I mean what do you mean by directly?

Guy Daniels, TelecomTV (10:30):
Well get involved in the AI visioning of AI infrastructure.

Yue Wang, China Telecom (10:33):
Okay, yeah, I see. Well I think from the perspective of China Telecom, we do have the a long-term vision of the network and the cloud and need to converge. So in the sense that in the past you provide the connectivity and you provide the connectivity for the data, the messages, the videos, et cetera. But now in addition to the network load you also have the AI workload. So how do you actually serve both of them simultaneously? And we are actually discussing one of the use cases is for example, the smart glasses where you would need not just the connectivity for the glasses but also some of the edge compute to be provided because the factor size and also because the limitation on the power of the glasses. We do see there is this convergence between the network and the cloud or even another or the connectivity and the compute to support some of the emerging new use cases that is mainly driven by AI and we think that's a great opportunity for the operators.

Guy Daniels, TelecomTV (11:55):
Great. And this leads me to my next question and Eric, I was going to come to you on this one. Which proof points most convincingly justify this investment in telco infrastructure When the backdrop to this is telco CapEx is not expected to rise very much, we're looking at projections that may be flat at best. So we've been talking about use cases here. Are there any proof points that come to mind?

Eric Van Vliet, Dell Technologies (12:25):
I think first of all it's the speed things are happening. I think one of the big proof points is the speed that things are evolving. If we're looking to technology, any technology, if you go back 50 years evolution of the certain technologies, well at a certain pace, I don't want to call it slow, but it was slower than this now and things are accelerating more and more because the world is now working together. Compared to 50 years ago people were working much more in isolation. I think one of the proof points is the speed of the AI adoption adoption in consumer's minds in education in the enterprise. I think all of us here use AI somehow while we work. We use AI when we're at home, our devices use ai, we don't even see it and that just gets bigger and bigger. So I think that's one thing.

(13:15):
And I think the second thing is that the adoption complexity is lower and lower. So small medium business, and I'm a firm leader that if you measure the small medium business adoption of certain technologies, and that's where the real proof point comes from as well because obviously large enterprises have sufficient funds to invest in people, but the moment things are really accessible for the small medium businesses to get better and to have the luxury of these new technologies and that's accelerating as well. So even small and medium businesses, they're using AI already today they didn't need to invest a lot of money to get started. They were able to do that pretty straightforward and make their business better. I think the risk we have in the telco industry, and obviously we don't work for telco, but the telco industry needs to be careful that AI technology doesn't become another over the top application.

(14:11):
I think that's the proof point that we have to invest and maybe also because the learning the telcos have done now looking at how on demand video and social media as an over the top application, it's very stressful for the network and not a real revenue generator for the network. So I think those things really show us that if we don't act now, probably it's going to in a couple of years you are going to be too far behind it. We're back where we were maybe 10 years ago or eight years ago behind over the top providers

Guy Daniels, TelecomTV (14:45):
And we absolutely don't want want that. We don't want to. No. Matt, do you see that the similar proof points justifying how you go about evolving your network technology, your infrastructure?

Matt Rees, Neos Networks (14:58):
I guess just pick it up on the OTT example Eric just gave for a second. I think it's almost inevitable that happens to an extent, a bit like the content caching that all of the big guys did when the video was driving all of the capacity demand. I think actually there's probably back to the early question, I think you asked UA about how can telcos participate in that more at a larger scale, maybe bigger than the uk. I think that all of those properties that telcos operate that are close to customer's edge become potentially good candidates for inference to be hosted. And there's a chance to sell the space to host inference close to end consumer. I think still in a world where CapEx then is constrained, how do we make sure that we structure our investments in the right way is probably partnering with the guys who have got the money.

(15:52):
That's probably the gcms, the content guys isn't it? And making sure that we are working hand in hand with them when they're working on their new big DC deployments. We're talking about what's the right model for bringing those dcs on net. And I've seen a few different examples of that, whether it's I'm asking the telco to fund the construction in some of those DC operators wanting to own that last mile into the network themselves and pick up an existing network footprint, but they want to own that last mile. But I guess what we have seen in most cases is they're all talking about four completely separate routes into a dc. So I guess there's definitely spades and grounds CapEx to be spent on putting that infrastructure in place.

Guy Daniels, TelecomTV (16:38):
Great. Well let's develop that one. I've got another talking point down here, so let's develop that new way. I'm going to come to you first because this participation in this AI data center sector potentially is huge. I mean if you like to believe the big consultancy firms, the numbers are astronomical in the trillions, whatever numbers you believe, they're big numbers. So how can telcos participate in this? And I think one of those big firms regards telcos is energizers, that's their category. They're in energizers along with power companies and what have you, the energizers. So how do we go about this?

Yue Wang, China Telecom (17:20):
So I mean we did build the AI data centers and we did observe it generates the new revenue. I mean it's big investment. It is really big investment. But with that you also get the revenue return. And I think one of the benefit of doing that is to really start to see with this one unified digital infrastructure with the network and the network, we traditionally doing very well. And then with the data center you're now able to generate the service on top, the new AI driven services and from that you are able to drive the new revenue stream. And so I think one of the numbers I can give is with AI services from last year, the year on year growth is like 20%. So with the IOT data center, et cetera. And on the other hand, I think I read one of the articles I was saying, they investment on gen AI is jumps like 76% every year.

(18:32):
I mean I don't think Teleco can actually match that scale of investment. So back to the point of the people who has no money, I think in the earlier stage when AI was developed, we didn't enter that ecosystem. I mean cloud hyperscalers were there, they built the whole ecosystem with ai. We kind of lost that opportunity I think. But now with this new wave of the emerging AI services, I thought maybe then we have this opportunity to actually build the ecosystem around this emerging AI technologies. And of course that is based on a solid AI infrastructure.

Guy Daniels, TelecomTV (19:21):
Yeah, absolutely. Thanks. And Eric, is this something that you see across telco partners or is it specific telcos with a lot of clout behind them are going to be the ones in a better position here?

Eric Van Vliet, Dell Technologies (19:38):
Well I think building a data center is a very complex activity and is probably not as complex as running a network. But it gets there, right? The power requirements, the cooling requirements, especially for these mega data centers run ai. It is a astronomical. So I think from a telco perspective, if the telco participates with the data center buildout, it's disaggregated data center methodologies distributed through the network because that's the value that telco can bring. Anybody that has enough money can build the data center and you need land building done. Sounds complicated, but that part is not, but nobody can build let's say 20,000 little daily center. It's easy. That's hard because that's 20,000 sites. But the telco scan, they're already doing that today with new technologies within the radio access network, it's server technology sitting there, streetside cabinets somewhere in the middle of nowhere, it doesn't matter.

(20:43):
Telcos can do that. So I think the value for the telcos within this AI compute data center build out is really how can they disaggregate that data center and drive the AI closer to the user. But that's really also where we want it to be because that's where the data's contained. And the other thing I think is when you do that, you also, and I know that we all in this industry are very aware of it, but the service provider is a trusted vendor in every country. Thank you. You trust I can call an ambulance, I can trust you with my data. Everything's in country is regulated, whereas other hyperscalers are not regulators, no regulation for cloud computing usually. So another benefit I think for telco service provider is that they are part of that. So I think it's also an ask to the regulators to drive that demand

Guy Daniels, TelecomTV (21:40):
And Yue, you were agreeing there with a disaggregated approach.

Yue Wang, China Telecom (21:43):
Yeah, yeah, precisely that's what we did. I think what we've observed is how traditionally AI training happens at the cloud you would have the centralized server with AI accelerators and you would perform the training at the cloud. But having centralized cloud actually has issues. I mean if you only think about the power supply, there's no way you can have that giant data center that deals with everything. And then there is also a significant trend where the compute and the inference is moved closer to the user and to the adjunct fire equipment. And I think what we did is the AI data centers one is to be able to perform this distributed AI training and that training can achieve according to our experiment, like 97% compared to the centralized training. So in terms of performance, that is already quite good. And then on top of that, with the compute closer to the user, you will be able to address the latency issues, the privacy issues. So yeah, that's where we see the opportunities are.

Guy Daniels, TelecomTV (22:56):
Great. And Matt, going back to your point when you said looking and partnering those who have the money and the experience, can you combine that with a disaggregated approach?

Matt Rees, Neos Networks (23:07):
Yeah, I think so. And I think, well, I guess the latency thing, and maybe you use market's a bit different to mine, a lot larger geography to cover than that uk. But I guess the consumer use cases that everyone's doing at home of generating the next latest funny image, there's no latency dependency there, right? No one is too fussed. They have to wait a little while for their image to pop up. But I think as enterprises start to largely adopt it, then that's where you're going to have to have that latency consideration when you want decisions made in near real time. Especially if going back to the prior panel around the objective AI may be making network control decision and fixing fault. You want that happening rapidly. So I can see even that going out into customer premise potentially at some point. Not just the street furniture and mast locations that the guys were talking about before.

Guy Daniels, TelecomTV (23:54):
And you both mentioned about driving data centers closer to the users. How, what's the strategy for doing that? How do you do that and how do you make that work? Is anyone any examples or best practices?

Eric Van Vliet, Dell Technologies (24:13):
Well, how to make that to work it obviously it's not an easy adoption because operating such a data center that is disaggregated is hard, but it's happening today on-prem right now. Enterprises are building little data centers on-prem in their premise. They don't like it. They have too many assets to manage. They need an IT person. There's lots of complications there and there always will go something wrong. So I think one of the things to start doing this is start looking at the current use cases and how we can cost optimize that for enterprises. So how the telco can take off some of those requirements. We already saw that with, we had software defined WAN making it easier to create foreign enterprise networks less reliable on other technologies so that it can be more flexible and universal pe there's cases where the telcos, they sell a managed service and they sell a couple of cores to the enterprise to do whatever they want on their little data center.

(25:18):
I think the step is bringing it into the network rather than having it on-prem because then also the telco becomes more sticky to this because right now at the end of the connectivity it's easy to swap, but if the moment it sits in the network, it'll help also the telco to become more sticky and start small. That's the other thing, there's no need to do it nationwide. There's no regulation that you have to offer everything nationwide if it's AI related. But you can start looking at just financial sector targets and then building that out with services. I think that's the other thing that's doable. But yeah, so it's a different approach. It's something that I think will come more and more in looking at how the network can be more valuable is what are the top priorities I believe for all telcos, how to get away from the flat rate, flat rate trap for value, how do we charge more? So I think that's probably the best way for me.

Guy Daniels, TelecomTV (26:19):
Thanks Eric and Yue, do you need scale? I was about to say do you need scale to do this? But Eric's just been saying that you can start small focus in small areas. So is this something applicable maybe from your experience to all operators?

Yue Wang, China Telecom (26:32):
Well yeah. One thing I learned after joining Channel Telecom is that things can scale up quite quickly. I mean yeah, it's different. So we shouldn't be comparing. Yeah, I remember last year in DTWI was saying one of the gene AI cases where we were deployed in one of the provinces. So I started to looking to how large that province is. It turns out it's kind of half of the size of the UK and there are like 31 provinces when you deploy internationally it's 31 provinces. So it's that sort of scale. And then back to what Eric just mentioned on the private network, what we did is there are different use cases where you can explore the mining we thought is one of the good ones because it does require, because it's a dangerous place to operate. So you would want to have the little autonomous vehicles and the robots to do the operation and then you want to have the private network to be able to control the fine motion of the robots. And this is something that was talked in the panel earlier as well as connected via. So we do see this is also a very good case where Teleco has a place because mobility is so important.

(28:16):
So I mean cases like this and I think private network is something you can immediately see the return and it's not that challenging to start.

Guy Daniels, TelecomTV (28:26):
Good, good. Thanks very much for those comments. I do want to move on to another area because that's around the area of resilience and how should telco's price and guarantee resilience? Because we're hearing that the cost of downtime, network downtime to enterprises can be astronomical. It can latest, where was this? I think this was in Forbes last month that were saying that it's about $9,000 a minute now for large enterprises it can certainly rack up in value. So Matt, is there any approaches here we can look at as we evolve our infrastructure to try and guarantee more pricing resilience?

Matt Rees, Neos Networks (29:11):
Sure. Well I think I mentioned that the greenfield dcs are all seem to be building in fiber separation four different routes into a DC location by default in my experience. But I think and how do we then make sure that as telcos, that we are not on the hook for a $9,000 a minute penalty? I guess we have some similar pressures on us today. I have some customers amongst my customer base who are emergency services, critical national infrastructure. I heard some of the previous panelists talking about that today. So we have similar pressure whether it's ability to dispatch emergency vehicles or a heavy penalty as a result of downtime. But I think what we have to do is make sure that we build in tiering into that resilience that the customer can differentiate between which of their use cases are top tier can never drop the emergency service scenario.

(30:04):
And I guess there'll be monetary equivalence of that versus ones that they're not business critical, they can have a downtime if need be and building a pricing tiering to support that. So I think we do that today offering even the openreach in the uk, the RO two product, you pay a premium to have that separation. Connecting new greenfield dcs, you can have a four times version. So if you're really that nervous, four times separation at the fiber layer and the equipment on top, again, we can make sure that you've got physical separation. Most of those greenfield DC seem to be wanting at least two network operators in, so they're thinking that way ahead. I think already that they want separation as many different layers of the network as they can get it. And then I think lastly you've got to make sure that you're protected the way you write your contracts, that there's a good shared risk that the customer's got some of that as well as the operator and you're working together to make sure that you're keeping the service up and available.

Guy Daniels, TelecomTV (31:07):
Great. And Eric, we have some tremendous, tremendous telecoms networks, really, really solid, robust, innovative networks. But can we do more around resilience?

Eric Van Vliet, Dell Technologies (31:19):
No, you can always do more, but is it cost effective? Yeah, that's the question. Yeah, I think now I dunno how telco should price and resilience, first of all on the $9,000 of downtime, it's not just around connectivity but could be caused by a lot of things. So when we deal with the larger enterprise or any enterprise, resilience is part of our conversation as well. And it starts with redundancy and equipment or redundancy within the equipment to a strategy around management of the equipment, monitoring of the equipment. We build software tools also to help and that includes also telecom service providers to aggregate telemetry so that AI can predict something goes wrong. And I think that's another thing to think about is it will go wrong and it's more be prepared, take action before it goes wrong, or at least when it goes wrong, be fast. Because if it's per minute, time is money and that's a challenge for everyone. But again, I think the telco has a great environment that they've built a Brazilian network because of regulatory. So they also have the experience what to do for resilience. So I think yes, it's a challenge for everyone in every industry being able to deal with resilience. But I think that Elco has done a pretty good job over the years building strong stable networks

Guy Daniels, TelecomTV (32:51):
As we've been talking about AI and supporting AI infrastructure and AI traffic that got any impact on the work we do and the approach we take to guaranteeing resilience?

Yue Wang, China Telecom (33:03):
Well, I think it's kind of a double edged sword, right? I mean on the one hand if the network is highly autonomous and when there's a downtime, presumably you would be able to quickly identify where the 40 is and be able to recover the network quicker. But then on the other hand you would worry because AI is still black box ish and you would worry what it has done to the network. But I think the approach we are taking is similar to what Matt just said. You can provide the redundancy, you can provide, isolate some part of the network if it's really, really, really important. You would have your IP part customer and otherwise you provide SOA, it is also network slicing driven by ai, but then you would need to have some sort of redundancy and some sort of resilience check on what is going to happen in the network.

Guy Daniels, TelecomTV (34:13):
Great. Thanks very much for that. Before we end this session, I did promise that I'd come to one of our online viewer questions because our online viewers have been sending Tony's extra shot program, a lot of questions and we're about to finish extra shot. So there's a couple that we didn't get in and there's one here that links to infrastructure also links to our green discussion. And I'd love to be able to just throw this in very quickly now. And this came from one of our viewers today who's saying, should telcos be a bit more aggressive on their use of renewables and liquid cooling, et cetera for their infrastructure? Should there be more effort placed on this? Matt, any thoughts, immediate thoughts on whether we should be doing a bit more on our sourcing of our energy?

Matt Rees, Neos Networks (35:01):
I think obviously I think there's always more to do. I think my experience Neos and prior to Neos, I worked Virgin Media, I know some of my former colleagues represented here today as well, but I think I know that telco is and has been deploying things like solar onto the roofs of their pops and dcs to try and use more renewable energy coming into to supply the network again. We've also seen the industry start talking about some of those dcs that we keep talking about standing up micro dcs, containerized micro dcs in the corner of a field that's got a full solar PV deployment. And then how feasible is it that to connect that back to good robust connected fiber so that it can form part of that ecosystem. So I think there's plenty of initiatives going on. I think there's absolutely loads of opportunities to do more in terms of the things like liquid cooling, I might be better off letting one of my colleagues talk to that sort of stuff as it's typically happening more in the compute world. I think, yeah,

Eric Van Vliet, Dell Technologies (35:59):
Like Matt said, we should always do more for the environment. We all live here, we share the planet. We should always do more. So first of all, the answer is yes, but it's difficult because talk about ai. AI uses so much power and so that's challenge one. Two, being able to send data over the air, what we're doing, I find it still quite impressive that it all works. Sometimes we should sit down and think it works, but it also uses a lot of power. So I think yes, there's some low hanging fruits to do, there's still little data centers running 24 7 on diesel fuel. I'm sure they're out there. Some of your networks,

Guy Daniels, TelecomTV (36:39):
Oh yeah,

Eric Van Vliet, Dell Technologies (36:40):
Sure they're out there. That's something to look at maybe at the start. Those are easy ones that cost, but it costs money. So the question is should you do more? Yes, regulation maybe, but it's hard. If it was easy they would already do it. So it's a question of why, what is the incentive? And if you drive it through regulation and we have other issues. So it's tough. But no, as a simple answer, yes they should do more. We all should do more, including Dell. We do everyday things around power. How do we reduce power of equipment of our own infrastructure that we use renewables from packaging, everything's now to also to our laptop designs so that yes, they deliver what they need but they're recyclable as well because there are these devices like phones, laptops, they are very, very hard to recycle. So again, we do these things as well as though and everybody needs to do their bit.

Guy Daniels, TelecomTV (37:42):
Great, thanks so much. And Yue, please help our viewer out with that question. Do we need to do more?

Yue Wang, China Telecom (37:49):
I mean certainly we need to do more and actually we have already applied cooling to the data center. So we were saying it is challenging to build the data center and we have learned on the way and cooling solution is one of them and now we are looking to monetize that solution. So yeah, I think the numbers, if you apply the station cooling technology developed in-house, it's probably saving on thousands of tons of the carbon emission. So yeah, it's already there. Great,

Guy Daniels, TelecomTV (38:30):
Great. And I'd just like to thank all our viewers online who sent in questions. We've got a lot of questions so I really do appreciate you sending in questions to this year's World Forum. Well look, we're out of time on this session. We need to start our next session now. But first we have an announcement about our charity pinball tournament, which has been running these two days in which ended about an hour or so ago. I'm informed that the charity tournament has now raised about almost 11,000 pounds for the National Autistic Society since we began a couple of years ago. So thank you everyone who donated these two days. I really, really do appreciate this.

(39:16):
And what's more, we were keeping scores and we have a winner. So I have a trophy here because the winner is still with us. And I'd just like to say, Eric, well done because it's you. Oh great. This is your life. There we go. Fantastic. Thanks very much indeed. There we go. Brilliant, brilliant. Well done. Fantastic. Thank you so much. Great, well thanks very much and thanks everyone for participating and supporting that to, it's really great. So I'm going to ask our guests, leave the stage now and we'll start our next session and our next guest come up please.