- Nokia says its Nvidia partnership "continues as planned," with AI-RAN now demonstrated in live commercial networks
- Analysts see no near-term hit to AI-RAN, but longer term is murkier
- Nokia’s Q3 growth came from optical and IP, not AI-RAN
Jensen Huang’s abrupt departure from Nvidia last week left the telecom industry wondering what happens to Nvidia’s 5G/6G network ambitions moving forward – and where does it leave partners like Nokia?
Nvidia’s reach into the telecom ecosystem is extensive, but its relationship with Nokia stands out, bolstered by a $1 billion Nvidia investment in 2025. The two have since moved AI deeper into 5G radio processing with trials and proofs of concepts aimed at bringing AI further into the RAN.
Fierce asked Nokia about its relationship with Nvidia and was told that their strategic partnership “continues as planned.” In fact, “our AI-RAN work together continues to build momentum, with AI-RAN now demonstrated in live commercial networks,” the company said in a statement.
No change in short term
In the near term, Huang’s resignation doesn’t alter Nvidia’s technical alliance with Nokia, analysts said, pointing to recent milestones like T-Mobile’s live-network AI-RAN call and Elisa’s advancements in Finland.
Nineteen customers are now working with Nokia on AI-RAN, with commercial release targeted for 2027, noted Roy Chua, founder and principal at AvidThink. “Those will likely progress regardless of who is CEO at Nvidia over the next few quarters,” he told Fierce.
However, Huang’s departure removes the visionary evangelist who personally championed telecom ecosystems and introduces some near-term strategic uncertainty under interim CEO Colette Kress, said HyperFRAME Research VP Ron Westfall.
“Consequently, the commercial burden of proving AI-RAN's return on investment shifts more heavily onto Nokia under CEO Justin Hotard, obliging Nokia to lean on its silicon-agnostic anyRAN architecture to reassure risk-averse MNOs that its edge-AI roadmaps are compelling regardless of Nvidia’s leadership transition,” Westfall said.
What about the broader implications for AI-RAN?
Huang's resignation doesn’t derail the broader AI-RAN ecosystem because the architecture is already institutionalized through Nvidia’s $1 billion investment in Nokia, commercial Aerial RAN Computer deployments and nearly 20 global operator engagements, including SoftBank’s AITRAS platform and T-Mobile’s live commercial AI-RAN call, Westfall said.
“Rather than relying on a single executive's evangelism, momentum is now anchored by top-tier operators and infrastructure vendors committed to 2027 commercial field trials across Japan, the U.S., and Europe, meaning development and live network testing will proceed uninterrupted,” he said.
However, because the interim CEO brings a financially disciplined CFO background and Nvidia’s permanent successor may scrutinize lower-margin, long-tail telecom bets more strictly than Huang did, ecosystem partners such as SoftBank and Nokia may face heightened pressure to prove swift commercial ROI and edge-inference monetization rather than relying more on Nvidia to subsidize the market vision, he added.
What are longer-term implications?
Chua said longer term, things are not so clear.
“Telecom at Nvidia was a Jensen bet from back in 2018/2019 and despite the lack of traction until recently around their Nvidia Aerial SDK/platform for RAN, he's been supporting and pushing their telco team,” first with former Nokia executive Soma Velayutham, who joined in late 2017, and then by adding Ronnie Vashista in 2020 and long-time AT&T veteran Chris Penrose shortly thereafter.
RAN is a $30+ billion-a-year market, with silicon a fraction of that, compared to Nvidia’s $215 billion revenue in fiscal 2026 and over $100 billion for the quarter closing Nov. 17, Chua said, adding that a new CEO could reasonably ask whether it's worth the attention.
“I'm sure Nokia will be watching closely as Nvidia figures out the current situation – new CEO or reinstatement of Jensen,” he said.
Is Nokia putting all its eggs in the Nvidia basket?
Westfall said no to this question, arguing Nokia is not relying too much on Nvidia.
“Although Nokia’s high-profile marketing alongside Nvidia might suggest an exclusive reliance on GPU architectures, Virgin Media O2’s (VMO2) November 2026 London AI-RAN trial – which achieved roughly 19% to 20% spectral efficiency gains using existing Marvell custom Layer 1 silicon rather than Nvidia hardware – proves that Nokia has not placed all its eggs in one Nvidia basket,” he said.
Beyond Marvell’s embedded ReefShark/OCTEON silicon footprint, Nokia’s silicon-agnostic anyRAN architecture actively supports Intel Xeon processors with vRAN Boost, AMD EPYC CPUs and Arm-based platforms across hybrid and cloud-native environments, Westfall said.
“This multi-silicon optionality shields Nokia from single-vendor supply chain risk and gives cost-conscious mobile operators a pragmatic upgrade path to deploy AI-driven radio optimization on installed custom ASICs or standard servers without immediately committing to GPU-centric rollouts,” he said.
Chua said he partially agrees with the statement that Nokia is betting the farm on Nvidia.
On the baseband roadmap, Nokia has said its current Marvell chips are the last custom silicon it will build, so it now depends on Nvidia’s roadmap, co-design opportunities, pricing and continued interest and investment. “That risk was always present, it’s just heightened now,” he said.
However, Chua offered two qualifiers. First, “Nokia as a company isn’t riding on AI-RAN. Q3 growth came from AI and Cloud (up 123%), which is mostly optical and IP. Radio Networks grew 10%, helped by deliveries pulled forward from Q4 and none of that is AI-RAN revenue yet,” said.
Second, “anyRAN still runs on Nokia’s purpose-built AirScale basebands and the partnership with Nvidia is nonexclusive. So the installed based isn’t stranded if the GPU plans slip,” he noted.
Qualcomm/SpaceX RAN vs. Nvidia/Nokia RAN models
Westfall pointed to a new emerging threat from SpaceX’s selection of Qualcomm for its terrestrial RAN architecture.
He said the divergence between the Qualcomm-SpaceX and Qualcomm-Viettel RAN models versus the AI-RAN Alliance architecture championed by Nvidia and Nokia represents a fundamental battle over the economics and design of next-generation mobile infrastructure.
“While the AI-RAN model seeks to convert cellular base stations into GPU-accelerated edge data centers, Qualcomm and Viettel directly challenge this thesis by offloading Layer 1 baseband tasks onto purpose-built ASICs, delivering advantageous power efficiency, lower upfront capital expenditures, and localized AI optimization without relying on unproven edge-inference hosting revenues,” he said.
Moreover, “Qualcomm’s collaboration with SpaceX on 3GPP NTN and direct-to-cell satellite connectivity threatens to absorb rural and wide-area coverage budgets, diminishing carrier incentives to deploy GPU-laden terrestrial edge sites outside dense urban cores,” he said.
By aligning SpaceX’s LEO satellite payloads directly with Qualcomm’s Snapdragon 5G Modem-RF systems across standard terrestrial cellular spectrum bands, “the alliance can eliminate the need for specialized satellite handsets or custom tower gateways, shifting wide-area network intelligence from capex edge servers into orbital infrastructure and consumer devices,” he said.
In addition, “by combining Viettel’s full-stack open RAN equipment and Qualcomm’s end-to-end silicon control with ubiquitous orbital coverage, this ecosystem can offer operators a lower-risk, financially disciplined path to 5G-Advanced and 6G modernization which can avoid GPU software lock-in,” he concluded.
So buckle up. The RAN just got a lot more interesting.
Read more stories about Nvidia and Nokia on Fierce Network:
Nvidia fallout: With Huang gone, what happens now?
Nvidia CEO Jensen Huang resigns amid affair reports
Nokia and Nvidia keep AI-RAN momentum going as they edge closer to commercial reality