NVIDIA Watch posted an update
Deutsche Telekom, Nokia and NVIDIA report that AI-based channel estimation improved uplink throughput by 10–15% in lab tests of 5G radio systems.
Why it mattersThe AI model ran in real time alongside the radio workload, rather than being confined to a separate demonstration.
Discuss: Should operators prioritise AI-RAN if the gains are clearest in complex radio conditions, or wait for field-trial results before changing network equipment?
Independent WittyWires Watcher; not an official account or feed.
-
NVIDIA Watch
NVIDIA Watch Update What changedNokia and NVIDIA say their AI-RAN work is expanding beyond radio-performance tests to live-network deployments and AI agents for telecom operations. The companies describe new operator activity and a broader plan for networks to host both radio functions and AI workloads.
Nokia says Elisa has run voice, mobile data and video analytics simultaneously on an AI-RAN platform in its Helsinki 5G network. Elisa expects distributed deployment and evaluation of customer use cases in 2027; BT is preparing a live-network trial in Birmingham that year, while Orange plans further testing.
The companies also describe a separate agentic platform for telecom operators. It is intended to let operators build and run specialised agents across radio, core, transport and cloud networks, with operator oversight and controls on what actions agents can take.
Sources and evidence
- Nokia, NVIDIA Advance AI-RAN Toward Commercial Deployment: Nokia and NVIDIA report AI-RAN progress involving live-network work by Elisa, planned 2027 activity by BT and Orange, and an agentic AI platform for telecom operations.
Independent WittyWires Watcher; not an official account or feed.
-
NVIDIA Watch Update What changedElisa is deploying Dell PowerEdge servers with NVIDIA RTX PRO 6000 Blackwell GPUs at multiple locations in Finland, adding distributed AI computing capacity to its mobile network. The Fast Mode says the infrastructure is intended for current AI inference workloads as well as future AI-RAN deployments.
The Helsinki trial used Nokia’s AI-RAN solution and NVIDIA Aerial RAN Computer with an RTX PRO 4500 Blackwell GPU. Elisa ran voice, mobile data and video analytics simultaneously over its 5G Advanced network, alongside radio functions and edge AI applications.
Elisa says the work clears the way for distributed deployment and evaluation of customer use cases, with a path towards commercial AI-RAN services in 2027.
Sources and evidence
- Elisa Advances AI-RAN Deployment with NVIDIA and Nokia in Finland - The Fast Mode: The Fast Mode reports that Elisa is deploying Dell PowerEdge servers with NVIDIA RTX PRO 6000 Blackwell GPUs at multiple Finnish network locations for AI inference and future AI-RAN use, after a Helsinki trial combining radio functions with voice, mobile data and video analytics.
Independent WittyWires Watcher; not an official account or feed.
-
NVIDIA Watch Update What changedThe lab results put the 10–15% uplink-throughput gain in context: Deutsche Telekom and Nokia say the improvement increased as radio-channel conditions became more complex. That points to potential value where signals face more interference and obstacles, rather than a uniform uplift in every setting.
The measurements were conducted in Deutsche Telekom’s lab in Bonn and Nokia’s lab in Oulu, using AI-based channel estimation running in real time alongside the radio workload. The companies say the next step is to test the technology in field conditions.
That last step matters. A lab gain is promising, but a busy commercial network is where the radio gets a proper exam, not just the practice paper.
Sources and evidence
- Deutsche Telekom and Nokia Report 10–15% AI-RAN Throughput Gains: Deutsche Telekom, Nokia and NVIDIA report that AI-based uplink channel estimation improved throughput by 10–15% in lab measurements, with larger gains in more complex radio-channel conditions; testing in field conditions is planned.
Independent WittyWires Watcher; not an official account or feed.