Guy Daniels, TelecomTV (00:13):
Hello, you are watching the Green Network Summit, part of our year-round DSP Leaders Coverage. I'm Guy Daniels, and today we're going to be investigating the application of AI to help improve the energy efficiency of telco networks. We're going to discuss what telcos are doing themselves and in partnership with the broader community. And joining me now is Nilmar Seccomandi David, who is director of Autonomous Networks and non-IT infrastructure with the global CTIO unit at Telefonica. Hello Nilmar. It's really good to see you. Thanks so much for taking part in our summit this year. Now we're going to look at a few examples of how AI and in particular gen AI can help with improving energy efficiency of networking infrastructure, as we said. So can we start with the planning and site development phase? How can AI help here?

Nilmar Seccomandi David, Telefonica (01:10):
Yeah, of course. I think we have many use cases where AI is helping us enhance energy efficiency. For example, in mobile network planning, we are leveraging AI to predict traffic with a three years view. Basically, this enable us to design sites that minimize power consumption by ensuring the site is right size for demand. This also help us to prevent over-provision and unnecessary energy usage. Additionally, with digital twin technologies, we can select the most efficient cooling system and power architecture according to site conditions. This is more focused on mobile network. I also have an example from data centers. We're also using digital twin also from planning process, but we are simulating cooling system behavior. Before installing new servers, AI helps to identify the best rack for each new server and avoid the need for additional cooling system, and of course optimizing energy consumption. In the other hand, mobile by, in mobile deployment phase, we are starting a trial to use computer vision. It's this used in a certain process to eliminate the need for manual certification from each order. AI use photos taken after the service is completed to verify that the equipment has been installed accordingly to the original design, including the tails such as cables, connectors, identification, and so on. In this case, we are not saving power, but we are reducing scope one and scope three emissions.

Guy Daniels, TelecomTV (03:12):
Great. I mean, there's some really innovative examples there, digital twins and computer vision and hope we're going to be hearing more about those over this coming year. If we move on to talk about network operations, what role does AI play in helping to reduce your power usage?

Nilmar Seccomandi David, Telefonica (03:33):
Definitely operations is the key area where AI can optimize energy consumption. For me, it's not just about the best technology, it's about how we operate that. The main use case is to optimize energy consumption based on traffic loads. Solutions like sleep deep modes can switch off network elements during low traffic hours. In other words, we can predict traffic and based on that, reduce the energy consumption without any quality issue. We are saving up to 30% by using AI solutions. Additionally, AI also can optimize cooling system based on thermal loads and of course weather conditions. In some of our operations, we are using 3D thermal map to identify cold spots and then reduce cooling inefficiencies, and finally, in maintenance, AI has been used to detect and resolve problems in our network network. As a result, we have fewer tickets dispatched to the field here. Also, we are reducing in scope one and scope three emissions,

Guy Daniels, TelecomTV (05:02):
Some really significant savings there. You mentioned it's very impressive in terms of monitoring all of this in terms of metrics and data. Does AI help Telefonica to better measure and understand its network energy usage?

Nilmar Seccomandi David, Telefonica (05:19):
Yes, it does. AI provide many energy insights by correlating consumption patterns with network performance. For example, most of our sites are installed with energy smart metering, generating millions of millions of data. Machine learning modules can analyze those data and compare with data from other sources, detecting anomalies and of course inefficiencies. Also, AI dashboards also can offer a real time visibility into energy usage, enabling us to implement optimization in cooling power plants and renewable sources. Definitely, we now have much more data metrics and KPIs, which is helping us to make informed decisions.

Guy Daniels, TelecomTV (06:23):
Well, Telefonica is obviously a leader in this area, but do we need a broader industry-wide AI-driven ecosystem to help with telco's energy efficiency, and if so, what kind of partnerships should telcos be looking to create?

Nilmar Seccomandi David, Telefonica (06:42):
Yeah, we need a collaboration. As industry, we are looking to collaborate with organizations such as three gpp, Etsy, and also vendors and energy companies. Personally, I'm co-chair of the sustainable focus group in modern alliance, and we are continuously discussing the potential of ai. We also have been involved in r and d with universities and ai, AI startups. I think we have many areas to leverage, including energy efficiency and AI by design in new technologies such as 6g. Also, not just as a sector, but as a citizens, we need to help to improve energy markets. We need a more smart network adapting energy customer needs to the energy generation.

Guy Daniels, TelecomTV (07:45):
Absolutely, and the best of luck with that, and I really hope the industry gets around these initiatives. There's a lot to do. A final question for you though. There's so much focus at the moment on AI and gen ai, but what about machine learning? Because we've been talking about machine learning for a long time now. Is ML perhaps a better route than AI for achieving some of our goals towards creating a smarter, more energy efficient network?

Nilmar Seccomandi David, Telefonica (08:12):
Well, I think AI and machine learning compliment each other in energy optimization. Machine learning is particularly effective for pattern recognition and automating decision, basically based it on historical data. AI and gene AI enables automation and energy ation, but gene AI is not cheap. Again, I think that they compliment each other. For me, AI use cases typically follow a three steps approach. First, understand the challenge. We need to identify inefficiencies in energy consumption. This is the challenge. Second, ensure data availability. For example, as I mentioned before, collect real data from a smart metering and third select the best AI technique. Could be machine learning, computer vision, GN ai, for example, use machine learning models to predict and optimize energy usage. I think this is the three steps approach. I think the main question has to be, how can I solve my challenge? It's not how can I use gene AI Technology is the enabler, not the end. I think this is my thoughts.

Guy Daniels, TelecomTV (09:35):
Yes, absolutely. Focus on what the actual requirements are and tackling the challenges. Well, it's been fascinating talking with you, Nilmar. Thank you very much for sharing your views with us today.

Nilmar Seccomandi David, Telefonica (09:46):
Thank you very much.