Clarence Reynolds, TelecomTV (00:05):
Artificial intelligence is rapidly transforming the telecommunications landscape, offering unprecedented opportunities for infrastructure optimization and greater operational efficiency. Michael Clegg, vice president and general manager for 5G and Edge at Supermicro, joins us to discuss the emerging AI strategies in the telecom sector. Michael, thank you for being with me today. And to start, what opportunities does AI present for telecom infrastructure and for operations?

Michael Clegg, Supermicro (00:35):
So as we see in the market, it's actually very broad in the way I always like to define it is you have AI in the telco, AI for the telco and AI by the telco. And so by what that I mean is in the telco is ai, telcos will use AI for their own internal operations and networks. AI for the telco is a little bit more on the enterprise side. This is now where with gen AI you get a big overlap. Telcos are themselves very large enterprises. So customer support, customer experience, maybe some of the internal coding processes. And then AI by the telco, which is the newer one. Can telcos offer AI services? And we can discuss that a little later in the talk today. Now on the same side, there's also an initiative. The AI Ran Alliance has done the same thing looking at the RAN in specifically. So you have AI in the ran, you have AI and the ran, and then you have AI on the ran. So same idea that AI being used to improve RAN operations. You have AI basically running on the RAN itself and then AI being combined with RAN services to offer edge services. You already have this equipment out at the edge. So we see these initiatives really across the board in many ways that telcos will be able to take advantage of ai.

Clarence Reynolds, TelecomTV (01:56):
So with Supermicro's customers, what AI initiatives are telcos actively implementing?

Michael Clegg, Supermicro (02:02):
So it's actually very broad. I mean in the industry as a whole, some of the industry forums we ourself work with, as I mentioned, you already have the AI ran alliance team that has been set up by a number of leading telcos, particularly SoftBank is prominent in that and have made some public announcements about what they're doing. Over there you have the Telco AI Alliance, which is a number of telcos have got together basically saying should they develop their own llms? Where should they go with telcos? Should they make a telco specific model? We belong also to the telecom info project. And in there you've got an AI ran working group, well AI, telco working group looking at AI applications. In a recent survey actually done by telecom tv, what we see is that most companies, telcos already have an AI team, they have a dedicated executive, they have some focus.

(02:54):
So I would say that telcos are already committed to this space. Now we've seen quite a few public announcements working with Nvidia. They've announced quite a few Southern AI initiatives where telcos are working with their governments like in Korea, in Singapore, but Tel with a radar in Indoc, Swisscom did an announcement. So you see this telcos really looking to take advantage. Now as we talk to people in some of these groups and as we meet with people, when you actually talk to the guys inside the telco, they go, this isn't new. We've been doing AI for a long time. What's brought AI to the public attention or focus now is generative ai, which is a different class of ai, but that's the one that is more broadly.

(03:53):
It's the one that individuals can go off and use on a website or use on their phones. So suddenly AI has a lot more visibility. So on the one hand, AI has actually been around for a long time, a decade or two and he's been used at some level. And on the other hand, you've got this sudden big new wave and it's all new and exciting and the next big thing, it's going to open a lot more opportunities. I would say probably the big difference is for me AI is a tool and it's primarily a productivity tool. With generative ai, it is much more unstructured data, it's much more human productivity. How do we improve how humans do things? The type of work that humans would typically do, giving them a tool to become much more efficient. And I think that's why it's a lot more in the public eye right now.

Clarence Reynolds, TelecomTV (04:47):
So Michael, all of this implementation has got to have a financial benefit. How can operators quantify and maximize ROI for these AI technologies?

Michael Clegg, Supermicro (04:59):
So I'm glad you put it that way, is ROI. So there's always two sides to the equation. So one will be essentially on the cost side of the equation and that's probably where some of the more immediate benefits come from. So a number of the things that I spoke about a few minutes ago really are in that side of the equation. So if you look at thinking of AI as a productivity tool, so if they can make the RAN more efficient and in the van we're looking at things like energy consumption. There's initiative of TCOs that zero signal, zero power. If we can make the operation of the RAN more efficient, improve its effectiveness, that's going to give you a gain over there. I think BTF come out public is that they've already seen a 10 to 15% gain in coding efficiency by using ai.

(05:46):
So that's another internal operation easy one we spoke about on the enterprise side, applying AI to customer experience chatbots is going to reduce their customer support costs. So I think we will immediately see benefits coming in on the cost side of the equation that it's just going to continue to make telcos more profitable and more effective in that way. The longer question is what do we do on the income side of the equation, the revenue? Are there new services that can be offered? One of the ones we've looked at fairly extensively together with Nvidia is AI factories and we've done some webinars around that. AI factories is this idea that telcos who are very well positioned to do sovereign ai, which has a large privacy aspect to it, a large amount of data, a large amount of connectivity, and already the trusted supplier to governments.

(06:41):
So now you're talking about very large scale systems. Super Micro recently built one of the biggest liquid cool data centers with a hundred thousand GPUs in. So you're really talking about these that are starting to do training and some of the telcos I mentioned earlier, telco Alliance as should they train their own telco LLM, should they build their own LLM? Because LLMs are like languages, they speak a certain dialect and by making one that's more telco specific, we can potentially get more productivity out of it. Now telcos will be big users of AI as we mentioned, and there's an opportunity for them to offer that as a service to the enterprises. They're very enterprise focused. There's an opportunity to offer platforms as a service. We definitely see early interest in GPU as a service, as many, many customers starting get into ai. They're starting just with basic GPU functionality that will then grow and then down the road into package services.

(07:37):
So that's on the core side of the equation. If you go out to the edge, the AI ran alliance is already, their thinking is there's this edge compute that you have to put in for the RAN operations anyway. Compute is getting more and more powerful. Once you create DU Pools, you're going to end up with this excess compute capacity, especially when the network utilization is low. Can we take some of that capacity and apply it to AI processing either for an internal service, so that's the AI in the RAN or for an external service that's AI on ran. And so there's a lot of edge latency use cases that naturally fit with the 5G and upcoming six G and edge compute and that are going to benefit from additional AI functionality. And we see in there some projections that by generating the money of the ai, you can, by sharing that compute with your van, you can significantly reduce your van costs.

(08:29):
So I think operators are and will explore different ways to get into the market. They will definitely benefit on the cost side of the equation. And we see through AI factories, we see through AI in the van, we see through Edge Compute the large GPUs as a service on the infra side. There's many opportunities for them and I think it's key for them to move a little bit away from. They always want to away from being a connectivity player to being more of a services play. And AI is this new class of services that allows them to enter on the new market area. Again,

Clarence Reynolds, TelecomTV (09:02):
Michael, thank you for sharing your perspectives today.

Michael Clegg, Supermicro (09:05):
Yes, thank you. I enjoyed the talk. It's very exciting times. We're going to see a lot of activity in the AI space, so looking forward to what comes next.