Guy Daniels, TelecomTV (00:12):
Hello, you are watching telecom tv. I'm Guy Daniels and in this program we will be exploring the ways in which telcos can benefit from the huge AI opportunity. And joining me to share his views is Yaming Wang, who is Senior Director of Edge and IOT Solutions at Supermicro. Yaming, We've seen that AI and especially generative AI, has already achieved huge momentum. So how are telcos taken advantage of the AI opportunities?

Yaming Wang, Supermicro (00:45):
So from our point of view, we have been working with a lot of customers. There are largely three categories of a use case. Number one is Telco can use AI to optimize its own internal operations. Number two is on the network side. The telcos can really bake AI natively into its network and largely the telcos can offer AI as a service to its customers.

Guy Daniels, TelecomTV (01:22):
Well let's talk about those three areas in more detail then starting with business. Can you give us some examples of telcos using AI for business operations?

Yaming Wang, Supermicro (01:31):
Sure. We all know the power of charge GPT. Imagine that operators has the tremendous amount of data from its subscribers. And with all those data, the operators can use that to optimize its customer experience by offer competitive products and also predict the future usage of the consumer trend. And also with products chatbot, they can offer better customer experiences when customer have questions or issues. So this is really changing the operators inside their operations a lot

Guy Daniels, TelecomTV (02:25):
And what Supermicro products are best suited to supporting this area.

Yaming Wang, Supermicro (02:31):
So if the operators want to use the products inside their existing data centers, they are large GPU servers such as HGX family or others, they can put into their data centers to provide the training and inferencing. And they are use cases where smaller GPU servers can be put into local data centers for improving customer experiences as well.

Guy Daniels, TelecomTV (03:06):
Okay. Well let's move on to the second area you mentioned. Can you tell me more about the use of AI in the network?

Yaming Wang, Supermicro (03:13):
Yeah. For example, we have been working with the operator that want to use the AI to help to identify the network of problems when that happens and also be able to solve the problems faster, which is today they're using some traditional tools, it's time consuming and sometimes it's not as accurate. Another category is really bake the AI into the network itself to use GPU servers as the fundamental building blocks for the future network, including some of the 5G and the six G as well.

Guy Daniels, TelecomTV (03:54):
I see. And again, let me ask you, which of your products are best suited to this area?

Yaming Wang, Supermicro (04:01):
Yes. Before I got to product, we recently joined the AI Ryan Alliance, which is revolutionary in the building AI into the network. So with that we've been offering MGX Grace Hopper based system to customers to build the AI run capabilities in 5G and into the future six G network. They're also PIE based GPU servers for far edge capabilities as well.

Guy Daniels, TelecomTV (04:35):
Well let's come on to the third area you mentioned earlier. Tell me more about AI based services.

Yaming Wang, Supermicro (04:41):
AI based services. They are two use cases I want to mention here. Number one is AI factory because the large language model, as we all know is for the booming of the rhythm. The AI demand really should be customized to its own country or regional optimization and things is also considered as critical. Infrastructure operators naturally can be the one to build the AI factory with optimized large language model for its own country and for its own region. So that a major use case. We have been working, for example, SK Telecom made an announcement with Supermicro and media and Lambda Labs to offer this in Korea. Another part is the AI services can be offered as a low latency to customers via the operator's network because operator has the last mile. So it can offer a faster response time without the large amount of data has always been uploaded to the cloud or data center to solve the inferencing problems.

Guy Daniels, TelecomTV (06:07):
So which of your solutions are best suited to support these use cases

Yaming Wang, Supermicro (06:12):
For AI factories? The large A TU or four UGPU servers are best suited because the high demand of computing powers for a large language model will require building many super clusters to support this capability for the AI inferencing on the edge typically require smaller servers such as MGX, grease Hopper servers or PCI based GPU servers with front IO and smaller form factor to fit into the environment.

Guy Daniels, TelecomTV (06:50):
I've got a final question for you. Yamen AI is regularly associated with requiring a high power demand, yet at the same time the telcos are engaged in a big push towards sustainability. So how do we reconcile these two?

Yaming Wang, Supermicro (07:05):
It's important to notice that AI is inevitable to power us into the future and with the demand of power consumption, we think liquid cooling is a critical piece in the solution. With the liquid cool, we will be able to put in a twice amount of servers into the same footprint. That will improve power efficiency and solve a lot of footprint restrictions as well. Supermicro has been leading in this area and we like to work with the telcos to put AI natively into their network.

Guy Daniels, TelecomTV (07:49):
Well that's great to hear. We must leave it there. Yaming, very good talking with you. And thank you so much for sharing your views with us today.

Yaming Wang, Supermicro (07:57):
Thank you for having me.