Telefónica looks to the role of automation and AI in sustainability

Over the last decade Telefónica has reduced total overall CO2 emissions by 49%, which reflects a 91% reduction in operational emissions (Scopes 1 and 2) and a 34% decrease in Scope 3 emissions.

Insight caught up with Nilmar Seccomandi, Director of Autonomous Network and Infrastructure in the Global CTIO Office, and Maya Ormazabal, Global Director of Sustainability at Telefónica, to discuss how the company is tackling CO2 emissions amid growing AI usage, and baking sustainability into future networks and AI deployments.

Like other telcos, Telefónica has benefited from huge, one-off infrastructure replacements, namely decommisioning its copper access network, or moving wireless traffic from 4G to 5G to reduce its Scope 1 and 2 CO2 footprint. Fiber broadband networks, points out Seccomandi, are 85% more energy efficient than copper, thanks to technologies such as XGS-PON. The company has also benefited from network virtualization, which is 30% more energy efficient than bare metal servers, says Seccomandi.

Automating energy effciency

The period of large network upgrades is largely over, leaving less low-hanging fruit. Now the company is looking at how operational automation, including the development of autonomous networks, and the use of systems such as digital twins, can help with the next phase of decarbonization.

Telefónica has a relatively long history in promoting sustainability, having established its climate change office 20 years ago. In addition to setting corporate climate targets, the office supports operating companies in Spain, Brazil and Germany in advancing their own initiatives.

Over time collaboration between technology and climate teams has deepened. Whereas synergies with technical initially focused on energy efficiency and renewable energy cources “now we have broadened the objectives, and we are working in a lot of different initiatives,” says Ormazabal. “We work very closely, not just to define our our targets, but also in the projects we are implementing,” agrees Seccomandi.

When it comes to autonomous networks, for example, Seccomandi emphasizes the importance of being guided by sustainability principles from design through to implementation.

“When we created our AN program one of our main targets was energy efficiency by design, sustainability by design,” says Seccomandi.

Indeed, out of the eight principal KPIs Telefonica has developed to measure autonomous network progress, two relate directly to sustainability, namely energy efficiency and improvement of its CO2 footprint.

AI pros and cons

However, AI, as Ormazabal points out, “offers opportunities to reduce the energy consumption and efficiency in the networks, but also has challenges related with the water consumption and energy consumption of data centers [that are] not affecting directly our own infrastructure ... and ... challenges related to traffic increase.”

Examples, of how AI helps save energy include one of Telefonica’s high value autonomous network (AN) scenarios, implemented by Vivo in Brazil, called the Fractal project. It uses artificial intelligence to perform data analysis that helps engineering and implementation teams optimize network capacity planning.

Vivo uses the system to autonomously identify the best expansion points and minimize commissioning and maintenance costs. The implementation highlights the strong link between reducing CO2 emissions and reducing the energy needed to run optimized networks.

measure the energy consumption per token for different AI models as part of a GSMA program. Again managing AI token usage reflects how energy effiency and cost control are intertwined.

“We have been implementing agents, some cases in production, in some cases in Minimum Viable Product or trial, we are realizing that the observability layer offer a view of the performance of the agents but also the tokens they are consuming,” according to Seccomandi. “It's very important because in some cases we are seeing that some agents are not so precise and they are consuming many more tokens than necessary. We are not just ... implement[ing] ... agents, but also analyz [ing] the performance KPIs to be efficient in both costs and sustainability.”

Other examples of the use of AI to inform more sustainable choices include a digital twin that shows via a thermal map where the cooling systems of Telefonica’s data centers are wasting energy and could apply power-saving features says Seccomandi.

Other initiatives, which do not necessarily rely on AI innovation, include the use of renewable energy, smart metering, or lithium batteries on what Seccomandi calls smart sites.

The Scope 3 data conundrum

Lowering Scope 3 emissions is a stubbornly difficult challenge for most businesses because they result from activities that lie beyond their direct control. They arise, for example, from the way customers use services, or how suppliers develop or deliver products and services. They also typically represent the biggest -- and in the case of some telcos, an expanding -- slice of the overall carbon footprint.

A 34% reduction of scope 3 emissions is therefore no small feat. Nonetheless, reductions in Scope 1 and 2 emissions means Scope 3 still represents more tham 90% of Telefonica’s total emissions.

The company is therefore working hard with suppliers and customers to further decrease emissions. CDP, which runs an independent environmental disclosure system for companies, capital markets, cities, states and regions, recently acknowledged Telefónica’s efforts and awarded the company the highest rating (A) in CDP’s Supplier Engagement Assessment (SEA) for the seventh year running. Telefónica and Telefónica Brasil are among the 40 global telecoms companies and 1,400 companies across all sectors on CDP’s SEA ‘A List’.

And Ormazabal says she is seeing greater collaboration with suppliers and “more specifics on ...lifecycle assessment information related with carbon emissions of specific products.”

In particular, “there is more action because we have better data, and because we are talking [a] ...common language, she says. “It's a really holistic approach. The most advanced suppliers and the most advanced companies have had the same challenges, so we have specific conversations that we find very, very useful.”

Further collaboration is still needed to drive more improvements.

“We are having more quality conversations with our suppliers,” explains. “If we have better information, more accurate information, and more collaboration within the whole sector, we will be able to reduce it.”

Here, she sees European regulation playing a helpful role, including when it comes to managing the impact of AI. “There are some pieces coming in Europe [that will help] ...everyone to make better decisions, and to choose whether we need to use AI or not, because it's not only about suppliers, it's also about customers,” says Ormazabal. “It's about improving transparency, giving more information, and raising awareness, and also putting [in place] the legal framework.”