Is AI making telcos smarter in tackling energy efficiency?

Communications service providers’ (CSPs) efforts to use less energy amid data traffic growth are bearing some fruit. Their average energy usage worldwide declined 1.1% in 2023, according to MTN Consulting.

However, in the same period energy usage by telcos’ webscale suppliers increased 16.4%, which promises to increase telcos’ indirect or Scope 3 emissions as they move more operations into third party clouds (see chart below).

Sustainability is not the only consideration. Energy can account for as much as a quarter of telcos' opex, and in recent years they are increasingly turning to AI to help them drive greater efficiencies.

However, it is still early days, and typically the use of AI is not systematic.

“Tier one operators are applying [AI], not massively, and probably not in substituting the human loop,” as Diego R Lopez, Senior Technology Expert, Telefónica and ETSI Fellow explained during Telecom TV’s Green Network summit. At Telefonica, for example, “we have AI systems that are being applied here and there for different activities that are mostly focused on assessing decisions that are taken by engineers and planners,” said Lopez.

However, that is changing. Currently, the company is working “to mix several different goals … [and] the coordination of different AI models,” according to Lopez, so that it can strike a balance between configuring networks to save energy and the need to guarantee network stability and high levels of user experience.

The greenest path

International operator Colt employs AI to address other issues it faces, including variations in energy prices across its multinational footprint.

The price per kilowatt per hour can double across borders, according to Mirko Voltolini VP of Innovation, Colt Technology during Telecom TV’s Green Network Summit.

Cost is therefore an important consideration when deciding where to route traffic. But it is not the only one. Its customers also have sustainability targets and the company has developed an interface that allows them to choose the greenest route.

Nonetheless, Voltolini admitted that a global network with hundreds it sites makes it “impossible to do optimal planning”. Therefore, Colt Technology is using AI for demand forecasting, which overlays a prediction of where traffic will go with information about where energy is the most efficient and greenest, he explained.

Indeed for most operators “many of the sites have been around for decades and they've been optimized for reasons,” other than energy, according to Beth Cohen, Telco Industry Analyst, Luth Computer Specialists, Inc. And because telcos have thousands of sites they have tended to have a diffuse notion of overall energy spend.

“We can use AI to optimize at the edge …[and] in the core, but I think it's going to provide more dividends at the edge,” she stated.

In contrast, data center operators operate relatively few, very large sites that they can position near to cheap or renewable energy sources, and have long focused on controlling energy costs. As a result their tools and methods are finding their way into the telco sector, according to Neil McRae, Chief Network Strategist, Juniper Networks.

Virtual benefits

As networks become more virtualized and cloud native there will be new opportunities to plan with energy efficiency in mind from the outset.

“There is the possibility of having a balance of where you run the functions and where you run AI, and you can plan for these to be well balanced and well-structured without making a significant or dedicated investment on dedicated hardware,” said Lopez. “For simple models that would help you to save energy at the edge … you probably you don't need big GPUs … but simply a well-trained model.”

But AI does not have all the answers. "The people who work in these sites day by day ...have probably got the juiciest knowledge about the site, what uses power, what are the challenges," pointed out McRae. "And that data is often overlooked because we're trying to pull it from systems. I really encourage people to go out and talk to experts in each of these locations, because they've got gold in their heads that if you extract it, you can make them, you can make a massive difference in the modeling that you do."