- Nokia and Nvidia are moving AI deeper into 5G radio processing as their partnership approaches the one-year mark
- Nokia reports 10–15% uplink throughput gains in early lab tests with Deutsche Telekom
- More than a dozen operators are involved in trials or proofs of concept, with Japan’s SoftBank planning a commercial-traffic trial in 2027
It’s been nearly a year since Nvidia pledged to invest $1 billion in Nokia and form a partnership that puts AI-RAN at the center of everything, and they’re still going strong.
Today, as GTC Berlin gets underway, Nokia announced next steps in how it’s bringing AI-native networks to commercial reality with the help of Nvidia: new capabilities to improve radio performance, enable real-time intelligence at the edge and extend governed AI agents across network operations.
Specifically, the companies are now applying AI within the 5G signal-processing chain, including channel estimation, beamforming, multi-user MIMO pairing and scheduling. In addition, Nokia says its AI-native anyRAN software, running on Nvidia Aerial RAN Compute, provides a software-defined foundation for integrating AI-powered radio functions that improve network capacity and performance.
All of this builds on the AI-RAN platform they announced in July, which is now being integrated into live networks and inching closer to commercial deployment.
Nokia, Nvidia look to 6G
According to the companies, they’re creating the real-time network foundation needed for physical-AI applications while giving operators a practical path to improve today’s 5G and 5G Advanced networks and evolve to AI-native 6G.
“The next era of connectivity will be defined by networks that do far more than connect: They will distribute intelligence, continuously optimize performance and enable entirely new AI-powered services, including the emerging world of physical AI,” said Pallavi Mahajan, chief technology and AI officer at Nokia, in a statement.
Mahajan added that Nokia, Nvidia and leading operators are putting AI into both the radio network and network operations and that should help carriers squeeze more capacity out of their spectrum today, support physical AI applications and pave the way for more autonomous 6G networks.
In addition, Nokia said there’s early evidence of performance gains thanks to its AI-RAN solution. For example, initial lab measurements by Nokia and Deutsche Telekom showed 10-15% improvements in typical uplink throughput using AI-based channel estimation.
How much progress has Nokia made in a year?
So, a year into their collaboration, what does Nokia have to show for itself, other than some mostly pre-commercial milestones?
Roy Chua, founder of AvidThink, said Nokia’s public milestones are pretty concrete. They include but are not limited to an OTA call with GPU-accelerated Layer 1 at T-Mobile’s lab in Bellevue, Washington, and functional tests with Indosat and SoftBank.
More than a dozen operators are in trials or proof-of-concept (PoC) deals, and SoftBank is set to do a trial with commercial traffic in Japan in 2027.
“Nokia hasn't shared field data yet, though they are claiming better than 20% spectral efficiency gains today, 50% by 2027 and over 100% by 2028. Those are fairly optimistic and aggressive numbers. Nokia's own commissioned study calls the 50% a best-case envelope that applies only in dense, heavily loaded TDD massive MIMO sectors,” Chua noted.
Other operators have made AI-RAN progress, including:
- Elisa: The Finnish operator ran voice, mobile data and video analytics simultaneously on the same AI-RAN platform over its 5G Advanced network, according to a separate press release today. The operator noted in its announcement that this trial "clears the way for distributed deployment, the evaluation of customer use cases and a path toward commercial AI-RAN services in 2027."
- Deutsche Telekom: DT is developing AI-based channel estimation as part of a broader set of AI-driven capabilities to improve network performance and efficiency, according to the Nvidia/Nokia press release.
- BT: The operator is planning a live AI-RAN trial in Birmingham in 2027.
- Orange: The operator is planning lab and live-network testing for AI-RAN.
What did the $1 billion investment do for Nokia? For Nvidia?
For Nokia, the money itself matters somewhat, but the halo effect provided by the Nvidia investment probably brought more value, he said, providing a couple bullet points:
- It gave Nokia a possible post-Marvell silicon roadmap. “Nokia now says its current Marvell chips are the last custom silicon it will build,” he said.
- A refreshed position for the equities market. “The stock went from Nvidia's $6.01 entry price to more than 2.5x at its May peak. It has since given back much of that, but it still sits roughly 50% above where Nvidia bought in,” he said.
However, Chua added one caveat: Nokia's "AI & Cloud" growth so far is mostly optical, which speaks to its Infinera bet and IP – not RAN. Nokia’s Radio Networks segment grew just 7% last quarter.
For Nvidia, its $1 billion investment in Nokia is a mere “rounding error” next to its OpenAI and Intel commitments, Chua said.
Still, “it brought credibility and expertise to its AI-RAN effort, which to date has seen limited uptake since their Aerial platform was first announced in late 2019,” he said. “Before this deal, Nvidia Aerial was largely a SoftBank showcase. Now it has a Tier 1 vendor and a real customer path.”
Read more about Nokia and AI-RAN on Fierce Network
Nokia’s AI-RAN pitch gets reality check in new study
Nokia CTO: AI-RAN lets operators move at software speed
Nokia AI-RAN claims are ambitious but analysts are skeptical