Ray Le Maistre, TelecomTV (00:11):
So I'm here today with Prashant from Intel and Darrell from Wind River, and we're talking today about AI Open ran and what the two companies are doing together. So Darrell, we're going to start with you. How is Wind River using AI to address the challenges that come with the increased complexity of 5G Networks?

Darrell Jordan-Smith, Wind River (00:33):
Really in three components. The first component is we're looking at machine learning in general, machine learning languages to collect information and data and telemetry across the network infrastructure. The second thing we're doing is we're looking at AI in general to do analytics on that data and then we're looking at languages that enable us to more naturally interface with that data. So we can actually do interesting things around network maintenance, predictive maintenance, managing traffic flows, controlling antenna, predicting particular issues in order to really drive cost and efficiencies across the entire network, which can be huge.

Ray Le Maistre, TelecomTV (01:18):
Okay, so lots of opportunities, and Prashant, what advice do you have for telcos as the industry looks to integrate those AI capabilities?

Prashant Agarwal, Intel (01:28):
Okay, I would like to make a couple of points. First, in order to deploy the AI and take the benefits of AI in ran operators need to really finalize their processing strategy. But I mean what platform they will be using to run these AI workloads.

(01:44):
Usually we think about the GPU as the processing for AI workloads, but not all the AI workloads are same. They have different requirements. For example, most of the RAN processing including AI ML solutions, it'll be running on the network edge because of their latency requirement and their real-time nature. Though GPUs are excellent, they're good for the AI processing because of their parallel computing capabilities, but because of their power consumption and complexity, they are not the best for the network edge deployment. Other thing I would like to say that most of the AI ran use cases are inferencing, which are much smaller workload and they don't require that high processing. So we have to be very careful about TCO implications, but there is a cost to run the AI and we do not want to use the expensive matters like LLMs to run the same results, which they can achieve by much less expensive matters.

(02:49):
So my advice for operators will be just go through your use cases, think about the latency requirements, power consumption and the performance, and then try implement the processing which is best suit for your use cases. We don't need to go with a hammer looking after for a nail. And that brings to me my second point. Do operators need to invest in a new hardware to implement ai? Of course not. CPUs along with the accelerator for AI are the best efficient platform to run the AI and workload. They provide you good KPIs and they provide you the good cost power envelope. Intel Jones's CPUs, they come with the inbuilt AI algorithms and we have already proven that the RAN use cases can be implemented on intake platform using CU and deals and operators can use the same hardware to implement the AI workloads on the ran. So my final advice will be think carefully about the additional cost hardware and complexity to procure new hardware and maybe start using the hardware which you already have or the Intel platform which you already have to implement the AI workloads.

Ray Le Maistre, TelecomTV (04:03):
Darrell, do you have any specific advice for telcos here?

Darrell Jordan-Smith, Wind River (04:07):
It's very similar to what was just said. You've got to look at the use cases. You don't want to over-engineer those use cases. You want to be able to incrementally improve your network operations to the point where you can actually see cost and efficiencies from that. And the other point is everything is done at the edge or the near edge typically in terms of what needs to occur there. So the telemetry is collected there and there's lots of diversity there in terms of systems and technology. And from a Wind River perspective, for over 40 years we've deployed technologies to the far edge. We have a lot of experience and understanding how to fine tune to get the best performance security in those environments. So working closely with Intel, we are able to bring that solution together end to end.

Ray Le Maistre, TelecomTV (04:56):
Okay. Now to build on what you were talking about earlier, what would you say is the role of AI in the telco network? Or perhaps what should its role be? Where can AI have the greatest impact and how would you quantify that?

Darrell Jordan-Smith, Wind River (05:13):
I think from our perspective, what we're seeing is the normal things I mentioned before in terms of the cost of operating a network, the cost of actually doing the improvements in doing preventative maintenance, the cost of actually using lower skilled individuals because you can use machine learning languages to be able to see what's going on and identify problems so you can actually get that problem to somebody who's more expert in being able to solve it. So those are interesting things there that look at the cost of network operations and improving service availability, which are very important topics. The other thing I think that we're seeing from an AI perspective that's particularly important is how do we protect the data that's also in the network. So it's actually an interesting role in actually securing that information. And telecommunications companies are trusted partners with all of our data. Everyone has a relationship with the telecommunications company. I'm seeing AI have a bigger impact there. And then the last thing is telecommunications companies are partnering with lots of other service providers that are delivering AI type services, leveraging a lot of the technology, the silicon that actually exists out there. And I'm seeing very innovative based business models that are beginning to sort of bubble up to the surface.

Ray Le Maistre, TelecomTV (06:29):
Now Prashant come back to you and talk specifically about Open ran. What are the biggest challenges that network operators face as they roll out and start to manage their open ran networks?

Prashant Agarwal, Intel (06:41):
Okay, that's a good question. So I would like to highlight two main points. One is the TCO and another is the integration. We have come a long way in terms of TCO because the operators have started using our latest platform. So our existing platform based on G four is providing double the capacity compared to our last platform. And it is the first generation platform with the integrated vRAN boost, which is also providing like 20% power efficiency. And now we are bringing a new platform in next year in 2025 GNRD, again, we are aiming to provide double the capacity. So with all this additional capacity, we are really hoping that most of these cell site configuration can be served by a single server. So in past operators were using multiple servers for the cell site configurations. Now if they can be all implemented on a single server, that's a big saving in terms of the TCO and also the power saving less server.

(07:40):
You will have the less power when come to the integration with the nature of the multi-vendor Open RAN operators have to put little more extra resources in terms of the integration compared to the single vendor traditional ran. But seeing that vendor community has come long way. Now we are working with vendors to provide preintegrated solutions, which can speed up the whole deployment process. For example, Intel solution builder ecosystem, we have hundreds of vendors. We work with them together and provide this pre integrated solution. Wind River is a premier partner there. We work Wind River very closely to provide the Preintegrated V solution, which has been already deployed across the globe.

Ray Le Maistre, TelecomTV (08:25):
Darrell, do you want to add to this?

Darrell Jordan-Smith, Wind River (08:27):
Well, I think what we're seeing through the deployments that we're involved with Intel, we're hitting the TCO targets that the clients have actually set out to want to achieve. In addition to that, it's the promise of being agile with the physical infrastructure as well. You can upgrade the physical infrastructure without having to replace all the software. So you can actually unlock more efficiencies over time and deliver more services in your network with a very similar footprint. So rather than having a vertically integrated platform, which is one way of actually doing it where you're locked into that for a longer period of time, you can have a lot more agility with Open ran. And that's really the promise that we're beginning to see and our clients now are beginning to see the benefits of that.

Ray Le Maistre, TelecomTV (09:12):
Finally, what should the industry expect next from the Wind River and Intel collaboration? Darrell, let's start with you.

Darrell Jordan-Smith, Wind River (09:19):
So from a Wind River perspective, as many people your viewers may know, we used to be part of Intel. So we have a very good working relationship. We understand each other very well, but the trust is between the two businesses is there. That enables us to innovate. So you'll see us do a lot more things with Gaudi, you'll see us doing a lot more things with Xon, other base platforms from Intel, and we'll actually build our products and services to make use of the silicon and get to market faster, innovate faster. So that's what you're going to see from Wind River

Ray Le Maistre, TelecomTV (09:53):
And Prashant.

Prashant Agarwal, Intel (09:54):
As I mentioned, Wind River is a premier partner for us. We have lot of calibration in terms of engineering and r and d resources. So in future, when we bring a new platform, we are working already with Wind River and other partners to, by the time we have the new platform, they already have the preintegrated solution. So what you will see, for example, with this newer platform, the GNRD, which we are bringing in 2025, we have already started working with Windriver and we are quite in Advantage stage. So by the time the platform will be released, we will already have a Preintegrated V solution which can be deployed by the operators.

Ray Le Maistre, TelecomTV (10:30):
Excellent. Well Darrell, Prashant, thanks so much for talking with us today and bringing us up to date on what the two companies are doing together in AI and open Ran. Thank you very much.

Darrell Jordan-Smith, Wind River (10:40):
Thank you.

Prashant Agarwal, Intel (10:40):
Thank you.