T-Mobile sees 40% efficiency gains in the last year

T-Mobile strongly supports an idea from Nvidia about adding AI capabilities to mobile network infrastructure. The concept suggests that future radio sites could use Nvidia's GPUs to handle AI processing. Instead of only running the wireless network, these locations could also act as small edge data centers for AI tasks.

This is a project that may have lots of potential, but carriers already rely on AI for their networks and it's all about efficiency gains and performance. Ankur Kapoor, T-Mobile's chief network officer, spoke recently to Light Reading and shared some interesting information.

T-Mobile already uses artificial intelligence in several parts of its network, and the company views these technologies as an important step toward the broader AI RAN vision promoted by Nvidia.

One example comes from disaster recovery situations. In such scenarios, the carrier's cell sites can automatically adjust their coverage to serve areas where connectivity is most needed, such as locations used by emergency responders.

The system relies on a self-organizing network powered by real-time AI and predictive data models that can redirect signals dynamically. According to Kapoor, the AI algorithms are integrated directly into the customized radio access network (RAN), allowing cell sites to modify antenna tilt and other technical parameters within minutes to improve coverage.

The operator has also introduced a feature known as service-aware RAN as part of its 5G Advanced rollout. While not strictly categorized as AI in RAN, the capability allows radio sites to allocate network resources based on the type of service customers are using.

What this means is that the network can recognize different applications and assign only the necessary capacity to each one.

Kapoor indicated that the feature has delivered roughly 40% efficiency gains since T-Mobile began deploying it about a year ago. The company says it was the first carrier in the US to roll out this capability nationwide, with support from its RAN suppliers, Ericsson and Nokia.

Live Translation (and more)

Some days ago I told you about T-Mobile's latest AI stunt: the carrier launched a voice translation service that relies on an AI agent. This allows real-time translation during calls made from the phone's native dialer when at least one participant is connected through T-Mobile's network.

T-Mobile believes its highly distributed infrastructure could support this approach. The operator runs more than 85,000 cell sites and around 100 core network locations across the US, which Kapoor said makes its network several times more distributed than typical public cloud infrastructure.