Gabriel Brown, Heavy Reading (00:01):
Hey Andres. So okay, this is the last one. The pleasure of the last session is we can be a little more free flowing. These guys are going to hate that, but don't worry about it. It'll be fine. We are also going to take questions and so forth as before. What I'd like to do really is to run through and I'll start with Andreas Mueller at Bosch. He's at the far end here. Introduce yourself and tell us how you're using Open Ran, why you're interested in it, and what does it do to make what you do in your business better?

Andreas Mueller, Bosch (00:36):
Right. Thanks a lot Gordon. So hello everybody. I'm Andreas. I'm with Bosch. So it's about industry verticals. We are an industry vertical and we are engaged in different vertical domains and automotive in manufacturing, of course building automation and other areas. And we've been looking for a long time already in let's say 5G in general, right? What it means for us, we see a lot of value in private 5G networks, for example, in manufacturing but also other areas. And at some point we also got interested in open run because we see that open run may address some of the challenges that we are facing. And at the same time, maybe the industry verticals can also maybe address some of the challenges that the open round community is facing. And that's why we then started also to try to understand a bit bit better and also to get our hands dirty. So we have an open round network up and running at our research campus close to Stuttgart. We are still testing, it's not a productive deployment yet, but we are testing how good it is, what the challenges are, and we are especially interested in the round intelligent controller, the eps, the ARBs and so on where we see quite some things maybe that we can do with that.

Gabriel Brown, Heavy Reading (01:54):
And Andreas, Bosch and you personally were kind of quite an important voice from the outside anyway in my opinion on 5G four industry verticals back, if we went back 5, 6, 7, 8 year, whatever point it started, what's the trigger then for taking it from 5G to open ran? Why has that become, why haven't you just settled on there's good systems available that we can use and do the job for us? Why do you want to get involved and be putting your hands on the tools as it were?

Andreas Mueller, Bosch (02:25):
Yeah, we are still exploring, right? So also other solutions of course are there and we are also exploring them. But we see also, I mean one challenge that we in general still have if you look at industrial 5G in manufacturing is the ecosystem ecosystem development. Because at the very beginning there were just solutions for a public networks and then maybe they were oversized and to expensive to complex and so on for what we wanted to do with them and with open round of course if it's more modular, if it's easier, also lower entry barriers for new vendors and so on. So we see this as a way that as a richer ecosystem of suppliers at the end. So that's one reason. And the other one is that the verticals are simply very heterogeneous. So also manufacturing is one vertical, but you have completely different requirements depending on whether you consider a car factory or semiconductor manufacturing or whatever.

(03:18):
The process industry, the chemical industry. So these are very different areas and it means we always need tailored solutions basically that satisfy the needs and the requirements of these different areas. So that means there's probably not a one size fits all, right? One solution that satisfies all the needs of the manufacturing industry. And again here, open run of course might help to accomplish this. And the last thing, what I've mentioned already, the run intelligent controller, which is maybe the most exciting thing simply because it's a component that we don't have in the classical which GVP only systems. And here we see and we envision that we for example, can bundle in future our use cases with dedicated RPS and potentially also our reps. So that can influence the behavior of a network in the way we need it. For example, handover management is a very simple example where we did our first own RPS and so on to just gain experience and see what we can do with it.

Gabriel Brown, Heavy Reading (04:16):
Okay, good stuff. I'm going to jump over Luca for a minute, go to Bertrand and then back to Luca. The reason for this, you'll see Bertrand at orange. Please introduce yourself, tell us about what you're doing and how open Ran is playing or may play into it in future.

Bertrand Rojat, Orange Events (04:32):
Yeah, so I guess, so Bertrand so I'm working with Orange specifically in a business unit called Orange Events. We do a technical production for big and sport events and one of our key accomplishment with the Olympics last year as you know, orange was the technical provider, the sole telecom provider for the Olympics. So we did a fiber communication, IP communication, mobile communication, and we had some very new requirements mainly from the media about very immersive real time image production. And so we thought about it on how we could respond to this kind of requirements and the answer was very, very clear. It was mobile, it was 5G. And so we decided to deploy on some key venues, the Olympic stadiums that de France the two biggest arena in Paris, LA De France swimming competition, the AK arena. And of course also as you know, I'm sure you saw it, the opening ceremony was on the river sand.

(05:57):
So we deployed private 5G because we wanted to have a smartphone camera on all the boats on the last but not least, we did that as well on the Marse marina for the sailing competition. So we had that big requirement. The only solution for us was to go mobile. And the thing about what was required, it was very demanding because it's first, it's real time zero delay. We're talking about 80, 90 millisecond delay, glass to glass, it is uplink. And as you know, all the commercial network are 80% downlink, 20% uplink. We had to do exactly the opposite. So it was also a big challenge. And also when we talk handover, the way we have to handle handover, we did have to guarantee at video layer the quality of the service. So to do that, the only option for us was to go very deep into the technology to be able to tune the network, to tune the service.

(07:10):
And on the open technology, open one was the only answer for us. So we went multi-vendor, the RAN technology was provided by Amert. We used a W2 S antenna system and the core network was a Cisco core network. And because we had a big handle on the technologies, on the different aspect of the technology, we were able to optimize it for these very, very specific use cases. It was very successful in term of deployment, as I said, we were able to have on the broadcast camera we're talking 50 megabit, continuous throughput, very low delay on the opening ceremony, 200 smartphones streaming at the same time during the ceremony. And so all the thing was possible also because it was fully virtualized on regular server because for us it was key to have a very integrated approach, very easy to move, to deploy, to reassemble, including on the marina. So for us, open ran was key for the success and that's how we're able to successfully deploy these technologies on it's becoming, I would say, a kind of defacto startup for us because on the sport event, community wireless is becoming the key technology.

Gabriel Brown, Heavy Reading (08:55):
Congratulations by the way. It was an amazing job. I mean a huge amount of work and I think the way you set it up, I guess you can use Olympics a little bit as a threshold or a marker of different eras. I know when it was in 2012 in London that was still today, that resonates because that got the live video streaming. You could do it before but that really industrialized it and made it something that is everyday usage. I guess now with the importance of big large scale live events, this mobility media production is really a new era.

Bertrand Rojat, Orange Events (09:33):
As you said, we made a demonstration that wireless is now mature for live production on, when I say wireless, I'm talking about wifi, I'm talking about 5G standalone on commercial network and private 5G. So it's a combination of all these technologies. We made a demonstration that it is very powerful, very efficient, and of course we'll continue to deploy these technologies in the event to come.

Gabriel Brown, Heavy Reading (10:00):
Yeah, okay, good stuff. So get ready with your questions quite soon. People I'd ask Bertran questions about this all day otherwise, but let's go to Luca. I've practiced the name of your company, but you're going to have to say it, I think Aalyria Technologies, Luca?

Luca Lodigiani, Aalyria Technologies (10:14):
Yes, Aalyria is correct. Aalyria . Okay. So my name is Luca Deja. I'm systems architect at Elyria Technologies for all of our non-traditional networks and direct to cell activities. So we're a bit of special case in the sense that we are like a venture backed startup and we came out of stealth mode roughly two and a half years ago as a spinoff of the Google Loon project. So carrying forward some of the technology that were developed under Google X and really the one of the centerpiece. So we have two divisions, one that does more like free space optic technologies with proper hardware, but space time, which is our flagship software platform, specializes in orchestration and control plane for very complex dynamic networks that span from the physical, the physics themselves, especially like aerospace networks, all the way to the application layer and focusing on end-to-end service assurance.

(11:12):
And so for us as we venture into non international networks, three GPP networks more in general on direct to sell the integration with or and using the OR primitives and sort of becoming part of the OR ecosystem made sense because our platform should be seen as more like an API platform that has a digital twin model and a number of ware engines and so builds the whole physical model of the network and then makes it available to APIs and then uses APIs to drive components of the network like for example, ground station components, radio access network and so on. And our APIs are already open source. In fact, even our data model is actually open source. So for us using the OR architecture and APIs was kind of a natural way to go. The reason why we in particular think that Oren has an important role to play in non-national networks is because there are extremely complex and dynamic networks to manage.

(12:12):
So they require a lot of complex API interaction with diverse set of elements and also a lot of holistic visibility across multiple components of the network. And especially in non-gestational constellations, they're very spectrum limited, they're very power limited. So it's management of an extremely tight and constrained resource, which is power spectrum capacity overall. And so the ability of the OR architecture to open up the radio access network and to open up control of large scale aspects of the radio access network as well allows us to optimize the usage of powering space of spectrum and the coordination for example with national networks as well. So we see the specialist applications especially of non-national network like hard QOS, like for public safety, disaster recovery applications like maximizing the spectrum usage efficiency as a key role for

Gabriel Brown, Heavy Reading (13:13):
Oren. So the common thread I'm hearing from three speakers so far is they had a kind of a problem or something they wanted to adapt for a special use case or something they wanted to work on. Before we come here, just Luca again, if I've got this right, someone coming from a mild adjacency at least to telecom and cellular, I know it is telecom, but it's slight adjacency to the mainstream. How did you find adopting ran specifications, architectures and ultimately if it comes to that code from the community, is it an easy to consume and access world? Is it

Luca Lodigiani, Aalyria Technologies (13:53):
I think from a specification perspective, there is, especially for the things that we are mostly interested in, which is fairly low level control, there is a good framework, there is space for doing more in the sense that now there is flexibility in the service models for example, to do a number of things. But we are already and we're expected we will have to do more in terms of developing additional service models that are more specific to satellite. There are some APIs that are simply very, very focused on the terrestrial portion of the ran on the genome B and so on. That will definitely need to be enabled. We would like to see more development in the area of accessing to the scheduler for example, or to some of the primitive in the physical layer. And we realize that from a macro brand sort of perspective, this could be a challenge, but that's why we think that the non-national world is a particular different economies enough and different dynamics enough in the ecosystem where this can actually work really well because it's a subset of very complex problems that need to be solved where essentially the importance is maximizing that efficient usage of that fin resource rather than catering to large scale and mass connectivity

Gabriel Brown, Heavy Reading (15:12):
Potentially once you're getting to that level potentially that's where an open source radio implementation might give you some extra flexibility I guess.

Luca Lodigiani, Aalyria Technologies (15:20):
Yeah, and we have actually started a work item together with Keysight in OR and Alliance right now is in the work group one to study the use cases for non-national metrics and identify some architectural aspects or some interface gaps or some specification gaps. And so we are pretty much of the idea that we want to do it in the OR space in the open space because as we open source our own, let's say non-standard APIs, for us the natural evolution is to an open ecosystem to focus on the complex parts of the problem, not the framework itself.

Gabriel Brown, Heavy Reading (15:54):
Okay, good stuff. Okay. No questions yet. Ye Ouyang from AsiaInfo, first of all, introduce yourself and maybe, I dunno, tell us what you're doing in ORAN for private networks and if you want reflect on anything you've heard.

Ye Ouyang, AsiaInfo (16:09):
Sure, thank you Mr. Brown. This is Dr O Yang from Asia Info. We are China's largest telecom software company. Actually our focus on the O ran, it's more for the private 5G market. Actually I prepared just three slides to quickly recap the progress that we have make to leverage the AI native O ran for the private 5G last year and Asia info collaborate with some of the telcos and some of the partners from the verticals. Can I go there?

Gabriel Brown, Heavy Reading (16:50):
Go for it. Sure, yeah, take the click.

Ye Ouyang, AsiaInfo (16:52):
Yeah, very happy to be here again. I remember last year I introduced some of the pilot studies of the O ran solutions for private sectors, but for last year, for 2024, we truly realized that the O ran solution coworking with the large language models to deploy such a total solution for some of the vertical customers in China. The title is Leverage AI native O ran for private 5G, but just want to have two keywords as a key takeaways for everyone here. The first is AI native, the second is private 5G the O ran should go private, more the OR should go alien, native, more intelligent, more. But this is very quick recap of the run evolution in the past almost 40 years. But if you take a look at the overall roadmap over here, most of the evolutions generation by generation is driven by by the general purpose technology rather than the right self.

(18:06):
The first general purpose technology is cloud, the second is virtualization, the third is now the artificial intelligence. So all these external forces are main factors to drive the evolution of the ran. That's the point that I'm trying to make over here. So ai, artificial intelligence generated ai, large language models should be the next wave of the GPT, the general purpose technology to drive the next wave of the run evolution. So the total or solution from Asia info is not making much difference from the other vendors. But the unique strength over here that we are making, the difference that we are making is the keyword on the upper right corner. The ISAC square as you may know, isac the itegrated sensing and communications, which should be the top one or two features in six G on the way we strengthened this word a little bit from ISAC to ISAC square.

(19:16):
Each of the letter indicates a unique feature here, integral of the sensing, artificial intelligence, communication and computing. What do we have done for the ISAC square? It's over here Last year we coworked with Intel and Alibaba Cloud to deploy the first large language model native, otherwise we call the large language model driven Isec square solution, which means in our CUDU we have a comprehensive ator to manage and re the resources, all the resources across the CPUs and GPUs. On top of it we are driving and we are running the 5G protocol stacks for communication service and we are also running the large language model services for some AI unread applications over here. The typical scenarios that we are working on is in a specific warmth, the wind farm and the nuclear plant scenario over here, the field testing engineers and the field operation engineers running into the nuclear plant that is very close the scenario there is no way for them to come out to seek further help for the maintenance and operation work.

(20:49):
So in here we could work with Alibaba Cloud. We had the first vendor to deploy the Alibaba's large language model, A distilled version Q wind seven B distilled version inside RCU and DU. And the engineers inside the nuclear plant and the engineers inside a specific warm plant can leverage their assistant co-pilots, the customer care assistant, co-pilots, and also the maintenance co-pilot to get the guidance for their operation and maintenance work inside their nuclear plant. And at the same time, the data service and voice service inside the nuclear plant is still going on. So we are running the communication and artificial intelligence service in parallel in the same CUDU for this specific scenario. The difference that we are making is the Rader that we deploy here is to make sure the dynamic orchestration and management across all the different computing resources, C-P-U-G-P-U and also the FPGA on the way and last, I don't want to advertise these numbers.

(22:09):
The point that I'm making for here is the O RAN should go private more and the O ran should go intelligent more. Why when we go to a specific private sectors customers over there, they don't mind if it splits 7.2 or not, but the O ran solution is the first stepping stone to realize their network modernization to support the basic network coverage functionalities when they upgrade their private network from wifi 4G to private 5G. But on top of it, most of the projects and opportunities get from private clients is not around network project instead of which is a complete data transformation project in which natural courage is the step one. And furthermore we have done a lot of the digital transformation applications for them and then the R and the AI native functionalities provided natively provided by O Run should be a key breaking through point over here for us to gain more opportunities from these vertical clients. So just the two points that making here the O run, we want to go private more the O run that we want to go intelligent more. That's it. Thank you.

Gabriel Brown, Heavy Reading (23:39):
Fabulous. Thank you. Ye Ouyang.

(23:45):
Okay, nice little transition. GPT there. General purpose technology from generic pre-trained transformer in a seamless manner. Yep, very impressive. Just before we move on to Matteo, what would you just give us big picture view on O Ran and open ran in China, what's the kind of feel for it? I think from the outside I kind of have the impression, obviously CHI with us, but there's kind of a coolness toward the approach and the big vendors don't really want it there. I could be completely wrong, but what give us a little bit of a feel for

Ye Ouyang, AsiaInfo (24:21):
Sure. Chilling is from the telco side, right? I'm from the vendor side. We may have different views but actually we stay with each other. Actually in China I just say Orion is not going very aggressively for public 5G, but for private 5G obviously it is going to be very great potential. Most of the private customer, private sectors customers, they don't mind otherwise they don't care if it's split 7.2 or not. That's number one. But basically they want to upgrade their network from legacy, either wifi or 4G to private 5G in here. The first step as a stepping stone for us to enter into this vertical client is to roll out the basic private network courage for them and then on top of it, that's the key point that we are making for the step two is the O Run architecture, the R itself, the XPP, the RPP architecture enable us to provide more AI related applications for the vertical clients. So as I indicated just now on the stage, most of the projects that we get from the verticals are not RAN project only, it's a complete digital transformation project. So on top of O Ran the Rick architecture enable us to provide more AI related services and capabilities to them. That's a key point to attract them.

(26:05):
Then in this case we can add on some other value added AI capabilities. Not only AI four ran it is we leverage AI to optimize our network itself. They don't care, they don't mind. But AI on RAN and AI and the RAN over here are very critical for them. AI and ran to help them to save the cost of the networking infrastructure. They don't want to burn the money to roll out and to build another IDC save the money for them. The AI on Ran is on top of the platform from Greek. We are able to deploy more and more AI capabilities like the large language models that we roll out for China. Nuclear, we like it very much.

Gabriel Brown, Heavy Reading (26:55):
Another good example of how Iran's enabled some other capability that wasn't necessarily so easy before. Matto, you've been immensely patient. I'm going to bring you in the last person to be introduced today. I think Matto Fiorini from Erickson. Tell us who you are, what you do, why open rank can help with private networks and reflect on anything else you might've heard.

Matteo Fiorani, Ericsson (27:19):
Yeah, thanks Gabriel. I have to admit I'm not as organized with slide and demos. It was very impressive. But I will try to do my best just speaking through, my name is Matteo Fiorani, I work for Ericsson. I work mostly with cloud run with Open Frontal and I try to give an update of what Ericsson is doing in open run specific in the context of verticals. So I start by saying that a lot of the work we are doing today in open run is actually focusing on macro networks and CSP. So I would say our traditional telco business, we are bringing out a lot of components during this year and the past years as well that would pay the key building blocks of building open run network and open run solutions. That includes, for example, our SMO that is called Ericson intelligence automation platform. That includes non realtime intelligent controller, no realtime Rick with an R one API on top.

(28:14):
And we have already integrated a number of Arabs both coming from Ericsson and coming also from other vendors as well. So that is one component. Other components that we are bringing in is open interfaces. So we release open frontal for a number of configuration these years but also interfaces to world open management like the O one and O2 that's still telco specific for the time being. But these, all these interfaces will become available during this year. And the last piece of course is our cloud and solution that these are fully cloud native networks that we are releasing and working on with some T one operators right now. So while during 2025 we focus a lot on macro networks and traditional telco business. All these components that I just mentioned would become in the future also very useful to build solutions for verticals and private networks.

(29:04):
As you mentioned, Gabrielle, it's very good spot for actually for open run going forward. I can mention a specific couple of cases that we have in mind on how we could apply our open run components into private networks. The first one is that we see that with the fact that co servers are improving in terms of their compute capabilities, we see the possibility of building solution that run on smaller server with small form factors that are relatively cheap that can be easily deployed inside enterprises for example, or not necessarily enterprises, it could be even retail stores, smaller venues. Those would actually be a perfect fit because they are easier to manage, they are closer to IT ways of working of course. And they're also of course quite cheap to be deployed into retail stores or small venues. And the fact is that right now you could see that some of these co server offered the compute capability to run a relatively significant number of cells and good radio configurations.

(30:08):
Maybe you can't run the latest massive MI generation on them, but you can still run for example your FTD bands, LT, and then R. You can maybe run even TDD bands on that IMO, but you could go for four T, four R for two T, two R and you can build excellent solution for private networks with these type of components with type of solution that are very tailored for these type of applications. So we see a large potential there for sure. Another area where we see that a lot of possibilities and opportunities for verticals in general with open is of course with the non realtime risk and the Arabs there we have specific business cases in mind. For example, what we are looking at is some cases for public safety that could become very interesting where you could have Arabs that are tailored for public safety.

(30:57):
So today we of course are able to differentiate for example first responders from other users that you can already do. You have tools in the networks that allow you to do that. You could use quality of service, you could do slices. What we think Arabs can add on top is that they can refine even further what you can do. For example, you could say that you have a firefighter being at the center of the action and you want to automatically recognize that that firefighters is closer to the real points and you want to prioritize that over other firefighters. So that is another level of granularity. So you are even more precise into selecting which type of, not only which type of user, but which specific user should have higher performance in a specific given point in time. Another example similar to that is that you may want to have an Arup that for example, free up your frequency there when you have an event, an emergency event for example. So this is still related to public safety. So you have a hurricane or you have a fire and then you want for example to free up some frequency layer in that specific region for the firefighters, for the first responders to be able to access the network and have priority. That could be another use case that we're looking at. So these are examples of things that we can do with open run, with open run technologies where we can differentiate and help verticals in a way that we were not able to do before.

Gabriel Brown, Heavy Reading (32:23):
And do you think with the verticals that, how much do you anticipate these end user groups want to get hands on with the tools and the code versus it's great capability but it still has to be packaged and supported and so forth?

Matteo Fiorani, Ericsson (32:37):
Yes, I mean there is always an aspect of how well the application work with the rest of the software because obviously every time we do, for example, changes in the network configuration, there could be unexpected consequences, but all idea is they try to decouple as much as possible the underlying platform from the application and intelligence on top. That goes, for example, for the ROPs case where we would see that we offer the platform with all the API and the control points for a developer to be able to develop as much independently as possible. Then testing verification in the end-to-end network. We always expect that has to be done to ensure that when you have an emergency situation and you want your to kick in, there will be no action mistakes or miscommunication in between the apps and the platform for example. So that is still very important. We expect that for example, we work with that with CSPs and potentially also if this is what goes more in an enterprise cases with the receiver of the solution.

Gabriel Brown, Heavy Reading (33:35):
So I'm going to see about a question or two for the last few minutes. Fellow over there. Thank you sir. I didn't know this is the point where if you have a little stretch, I'm jumping on you for a question.

Audience Question #1 (33:54):
Alistair Burr, university of York, uk, there've been several cases use cases here where it looks rather like the front hall is not likely to be provided over fiber. I'm thinking of some cases in the summer Olympics. How easy is that? Is that a problem? And am I right in surmising that this wasn't always over fiber?

Matteo Fiorani, Ericsson (34:29):
I dunno if there is any take, I can maybe start

Gabriel Brown, Heavy Reading (34:32):
The way I understood the question. Part of it is, is there a requirement, is there a need for wireless fronthaul essentially?

Ye Ouyang, AsiaInfo (34:40):
Yes.

Gabriel Brown, Heavy Reading (34:41):
Yeah, I mean generally speaking it hasn't really, there's been a lot of talk hasn't there and it hasn't really ever turned out to be.

Audience Question #1 (34:49):
I was interested

Gabriel Brown, Heavy Reading (34:51):
Let, let's see, has anyone fancy?

Matteo Fiorani, Ericsson (34:54):
I can try to start. I mean I would say that it's a bit challenging to run frontal over a non fiber medium. It is technically possible for certain configurations, so for smaller radio configurations. So if you would run a massive IMO, I wouldn't recommend to do that. But for a smaller radio configuration that could be technically possible. I think that it's probably, we are not there to be honest. The latency requirements are still very, very tough to meet over a wireless or a non optical fiber cable. What could change over time is if IP for example becomes a more common solution for building frontal networks today we mostly run on ethernet, right? So that has different type of requirements, but if we would go more aggressively into IP-based frontal in the future, that could be a way to get around the need for fibers. But I would say that as of today would be quite difficult maybe with some specific configuration could be made to work. But I would say that with an IP-based frontal, that could be the way to actually enable it.

Bertrand Rojat, Orange Events (36:00):
Maybe to add to it for the Olympics on the marina in mar. So we deployed actually a 5G coverage on the sea. So all the antenna and BBOs were on boats covering the field of play and the core network was on the shore and on the back haul between the radio system and the core network was done through a wifi back haul solution. So a point to point bike haul that was optimized for it, it's about 150 megabit throughput, very stable. It took us a lot of time to make it work, believe me, it was a real challenge but we were very successful in doing it so it's possible in very specific condition to make it work. So that was backhaul, not fronthaul, it was backhaul backhaul between bbu. And are you on the core network

Gabriel Brown, Heavy Reading (37:08):
On boats though, right?

Bertrand Rojat, Orange Events (37:09):
Yes.

Gabriel Brown, Heavy Reading (37:10):
So that is immediately extremely difficult.

Bertrand Rojat, Orange Events (37:14):
We like the changes,

Luca Lodigiani, Aalyria Technologies (37:16):
So perhaps I got to give you an example that is slightly exotic compared to maybe the Olympics case and so on. But in satre networks and international networks in particular, there has been now a big discussion of what is the payload architecture going to look like? And right now a lot of the systems are either fully transparent, meaning that basically the air interface, the UU air interface goes all the way from the gateway feeder link to the satellite and then down to the user link with a frequency conversion, which by itself is a relatively new concept that doesn't have a lot of precedence in terrestrial deployments and has its own challenges. And this is actually one of the reasons why we think that high layer control plane that has an end-to-end visibility is important because the channel state information, for example, for the feeder link is in some ways completely opaque to both the RAN and the user itself because it's in a different frequency.

(38:14):
It might be KB band, QV band for let's say A UHF or LS band user link. Now there has also been discussion of, hey, what happens? Like could be an architecture where we actually put the radio unit on board and in that case classic CPR by and large it just doesn't cut it. So depending on what is the beam bandwidth product that the actual satellite can deliver, meaning what is the beam hopping architecture, what the scheduled architecture looks like and what is the carrier usage architecture. Actually the split 7.2 could be a very viable way to go because also thanks to the enhancements in non-national network release 17 where we can relax the hark processing limits, for example, some of the other physical layer timing aspects, we can actually put the low end of the physical layer on the satellite and we can manage the bandwidth usage of the feeder link this way by simplifying also a lot of the problems that come with a fully transparent feeder link allowing multiplexing and network sharing and so on.

(39:15):
The problem is that if we look at the current architecture of the ran, if we go for example outside of the full genome be on board and we start thinking about middle way architectures today, the problem is more about the architecture itself of the, like the current splits, some of the ones that the satellite industry is looking to try to solve. Some of the complexity like potentially DU on board with F1 split have their own challenges. So I think what we are expecting is that in six G, if we do consider non-national networks from the beginning and with more gearing towards regenerative architectures, even if we don't just go for the fully regenerative genome on be on board, we may have to find some trade offs to make the architecture a bit more conducive. Because the problem is that the over wireless links, either you are very bandwidth constrained or the link is heavily affected. Like if you're using to deliver high capacity, you have to use feeder links in QV band or even optical free space. And so there's a number of interesting challenges that come up with the fronthaul and middle hall transport

Gabriel Brown, Heavy Reading (40:35):
Gentlemen in front, this could be our last question, Andreas, if you want to comment, you're going to have to just jump yourself in the conversation here. Please Go ahead.

Audience Question #2 (40:44):
So question for a year, right? You mentioned the AI and run use case and you mentioned also cost saving was a major factor there. Just wonder if you could elaborate a little bit about where does the cost saving come from? Is that from the server that's used for both AI and application and run application or somewhere else?

Ye Ouyang, AsiaInfo (41:06):
Yeah, obviously this is a very good question actually. I introduced a similar use case in the P-P-I-G-T-I submit over there a couple of hours ago. Yes, if you want to do a before and after comparison over here, apple with apple. So if you don't have AI native inside the RAN side, what you want to do is you roll out the network either through or through CGPP and for the AI computing, what you will do is you build up your A IDC for every vertical customers. If they are rich, they have budgets, they build buildup, rollouts, their own A IDC, otherwise they will go with AWS Google Cloud or Microsoft Azure. And in this case the hardware along with the software capabilities of artificial intelligence are co-located inside the same network architecture. So we call it ai, AI native network over here. So it's very similar to the new concept championed by the A RL alliance here.

(42:24):
I'm not sure if it's a good location to mention something new in AI lines over there. So the first point is AI and ran over here it is the infrastructure shared across the AI and the network. That's the number one saving that we are making here to save the cost. The vertical customers, they don't want to build extra money to buy the computing powers from the cloud operators or building up their A IDC center over here. And the number two points that we are saving the cost is the AI applications integration cost over here. So consider, I have a very typical use case over here. If you, let's say any given vertical customers to have a video conference call service in office A and office B, they pay for different silos. They pay for the benefits from telcos, they pay for the video conference call software from WebEx, from teams, from tension meeting over here. They pay for the value added application cost for the meeting minutes service and so on. And right now, if you just deploy your video conference call service along with a large language model inside your CU and BBU with such AI and a network shared infrastructure, that is all the four silos come to one single solution, which significantly seep their cost. That's the AI native based overall video conference call service for our vertical customers. It's very popular for them.

Audience Question #2 (44:12):
Thank you.

Gabriel Brown, Heavy Reading (44:14):
And with that, thank you. I think that actually brings us to the end of the panel session and also the end of the event. So there's just one thing before we go. They're going to put on the screen a QR code with a link to shows audio around demos around Mobile World Congress. There's a map and all that kind of thing. There you go. So just as we close and leave the room, please join me and thanking the panelists. Thank you guys. Thank you.