Nokia's Open Fronthaul Gamble: The uRAN Move That Could Finally Make Open RAN Pay

Nokia’s Open Fronthaul Gamble: The uRAN Move That Could Finally Make Open RAN Pay

Nokia is supporting a new open RAN interface it hopes will make its offer look far more competitive in the 5G battle against Ericsson and Samsung.

That single line, buried in Light Reading’s coverage this week, is the most consequential open RAN story of the past three days—and almost everyone is going to miss it because the headline is about Nokia’s “5G revival bid” and the subtext is about AI-RAN. It shouldn’t be. The real story here is not that Nokia is putting new open RAN technology into its portfolio. The real story is which technology, why now, and what it says about the strategic trap that Ericsson and Samsung are walking into by demurring.

Let me be blunt: for five years, the open RAN movement has been selling disaggregation as a moral good. It hasn’t worked as a commercial argument. What works is a vendor being able to tell an operator: “You can keep your existing radios, you can mix our baseband with someone else’s, and you won’t have to rip out your footprint to get the next capacity bump.” That is an economic argument, not an ideological one. And that, finally, is what Nokia appears to be making.

What Nokia actually did—and why the framing matters

According to Light Reading, Nokia is throwing its weight behind a new open RAN interface with the explicit goal of making its offer look far more competitive against Ericsson and Samsung in 5G. The article notes that Ericsson, by contrast, “demurs.” That word is doing a lot of work. Demurring is not the same as rejecting. It’s the posture of a vendor that has a large installed base, a healthy RAN share, and very little incentive to make it easier for an operator to swap out its baseband while keeping its radios.

This is the classic incumbent’s dilemma in a standards-driven market. If you are Ericsson or Samsung and you own a big chunk of the RAN footprint, openness is a tax on your incumbency. Every interface you standardize is a door you leave unlocked. So you support open RAN in press releases, you join the alliance, you show up at the demos—and then you make sure the interfaces that matter most to lock-in, the ones between the radio and the baseband, stay proprietary or stay “open” in name only.

Nokia has been on the wrong side of that math for a while. Its RAN share has been under pressure, and its AI-RAN narrative—however loud—has been built largely on a bet that many observers regard as a massive gamble on Nvidia GPUs. Light Reading’s piece is explicit about this: the GPU bet “probably explains why Nokia has been so keen to advertise telco interest in its forthcoming AI-RAN offers.” When your differentiation is a compute architecture that your competitors can also buy, you need a second differentiator. Openness is the obvious candidate.

The AI-RAN story is real, but it is not the whole story

Let’s give Nokia its due on AI-RAN before we get to the harder questions. The company has been running proof-of-concept and field trials of AI-RAN using the Nvidia Aerial RAN Computer across North America, Europe, Asia-Pacific, and the Middle East. It claims a 20 percent enhancement in spectral efficiency during testing, with additional software advances slated for 2027 and 2028. At MWC 2026 it launched what it calls the industry’s first commercial AI-RAN platform, built on its anyRAN software and Nvidia’s Aerial platform, targeting more than 100 percent spectral efficiency gains by 2028. It has demonstrated GPU-accelerated AI-RAN with BT, Elisa, Vodafone Group, and NTT DOCOMO, validating that AI inference workloads and radio processing can run concurrently on shared infrastructure.

That is a credible body of work. It is also, strategically, a commodity path. Nvidia’s Aerial is a framework any CUDA-literate developer can take and tinker with. The chipmaker has positioned GPUs as a general-purpose substitute for custom silicon in the RAN, and it has assembled its own RAN software stack. If the compute layer commoditizes, then the vendor that wins is the one that makes integration, migration, and multi-vendor operation cheapest. That is an openness play, not a GPU play.

And here’s the uncomfortable part for the AI-RAN hype machine: Nokia’s own CEO, Justin Hotard, reportedly reeled off the names of operators giving AI-RAN a test drive—19 signed up for “initial deployments”—but the three companies he identified are less well known, and none is a telco. That is not a knock on Nokia; it is a reality check on the entire category. The AI-RAN story is moving from lab to field trial, but the commercial deployments are still thin, and the logos that matter most are still in “trial” or “PoC” language.

The Cohere angle: where the openness argument gets teeth

If you want to understand why an open interface matters more than another spectral-efficiency claim, look at Nokia’s partnership with Cohere. Cohere’s technology—endorsed after field trials by Vodafone, Australia’s Telstra, and Bell Canada, all three of which have now invested in the startup—is aimed at doubling spectral efficiency by 2028 on much of the existing infrastructure. Cohere CEO Ray Dolan has identified Nokia as the vendor partner during trials in the Bell Canada network.

Those trials, per the reporting, used a common, off-the-shelf server that can be deployed next to a basestation at a mast site to host Cohere’s technology. But a dApp—a distributed application—would be closer to a direct integration and probably come with service benefits. Dolan’s claim, made on a Telecoms.com podcast, is that this would achieve what xApps did not.

Read that again. What xApps did not achieve. The xApp was supposed to be the killer app of open RAN—the third-party application that runs on the RAN Intelligent Controller and unlocks innovation. In practice, xApps have been constrained by interfaces, by near-real-time latency budgets, and by the fact that the most valuable control loops stayed inside the vendor’s box. If Cohere’s dApp model actually delivers what xApps promised, it is not an incremental improvement. It is the commercial vindication of the disaggregation thesis—and it only works if the interfaces underneath it are genuinely open.

That is why Nokia’s new open RAN interface support is a bigger deal than another AI-RAN press release. It is the plumbing that makes the dApp economy possible.

Why Ericsson’s demurral is a strategic bet, not a mistake

It is tempting to frame Ericsson’s position as short-sighted. It isn’t. It is a rational bet on the shape of the next RAN cycle.

Ericsson’s calculus is straightforward: if AI-RAN capacity gains are real, the operator’s decision is not “which vendor has the most open interface” but “which vendor can deliver the capacity with the least integration risk.” Ericsson has spent years arguing, implicitly and sometimes explicitly, that openness is a means, not an end—and that the end is performance per watt, per euro, per site. If it can hit the spectral-efficiency and energy targets with a tightly integrated stack, it can keep its share and its margins.

Samsung is in a similar position, though with a different footprint. It has been expanding its private network portfolio across manufacturing, finance, construction, and IT, spanning RAN equipment, virtualized RAN, Open RAN, AI RAN, cloud-native cores, and private network solutions. Samsung’s openness is real but selective—it is open where it helps it win enterprise deals, and integrated where it protects margin.

The risk for both is timing. If the AI-RAN capacity gains arrive on schedule—2027, 2028—and if operators conclude that the fastest path to those gains is a multi-vendor, GPU-accelerated, dApp-enabled architecture, then the vendor with the most credible open interface wins the upgrade cycle. If the gains slip, or if operators decide that integration risk outweighs the theoretical benefits of openness, Ericsson and Samsung win by standing still.

Nokia is betting on the first scenario. Ericsson is betting on the second. Both are placing real money on the table.

The India and Korea signals: this is now a global standards race

Two other stories from the past three days deserve to be read alongside the Nokia news, because they tell you where the market is heading.

First, India. The Economic Times published a primer on 10 terms to know ahead of India Mobile Congress 2026, running October 7–10 at Yashobhoomi in New Delhi. AI-RAN, Intelligent RAN, and AI-native telecom networks top the list. IMC is billed as Asia’s largest digital technology forum for telecom, media, and technology, and this year’s agenda includes AI-native networks, RAN, autonomous and zero-touch networks, edge and cloud AI infrastructure, and network APIs. When a mass-market business daily in the world’s second-largest telecom market starts defining AI-RAN for its readers, the category has crossed from vendor marketing into mainstream operator planning. That matters for open RAN because India is where the economics of RAN are most brutally tested. If open interfaces can’t win in India, they can’t win anywhere.

Second, Korea. The Seoul Economic Daily reports that Korean carriers are racing to secure overseas patents in 6G, quantum, and edge computing. SK Telecom filed a PCT international application for technology that links radio access networks with distributed edge data centers to allocate AI computing workloads efficiently—distributing AI computation to a suitable location based on network conditions across multiple edge data centers. SK Telecom is also feeding research into global standards discussions alongside its AI-RAN field trials. KT has obtained US and European patents covering quantum key distribution. LG Uplus is jointly developing low-earth-orbit satellite network technology with KAIST.

The patent race is the tell. Global patent analytics firm LexisNexis IP estimates the 5G patent licensing market at about $15 billion a year. Carriers and equipment vendors are competing to patent candidate technologies—AI-based network control, RAN, edge computing—before standards are finalized. Whoever owns the interfaces that get standardized owns a licensing market. That is the real prize in the open RAN interface debate, and it is why Nokia’s move is not just a product decision. It is a standards play.

The hard questions Nokia still has to answer

I am not ready to declare Nokia the winner of this cycle. There are three unresolved questions.

First, is the new interface actually open, or is it “open” in the way that some vendor-led initiatives are open—published, documented, and effectively controlled by the founding vendor? Light Reading’s framing—Nokia is “supporting” a new open RAN interface—leaves room for both readings. The proof will be in the governance, the licensing terms, and whether a second vendor can implement it without Nokia’s blessing.

Second, does the AI-RAN capacity story hold up in the field? A 20 percent spectral-efficiency gain in trials is meaningful but not transformational. The 100 percent-plus target by 2028 is the number that matters, and it is a target, not a result. Nokia’s own leadership has pointed to 2026 as the first year of commercial trials for 6G, with commercial release targeted for next year—an aggressive timeline that will be tested by physics, economics, and operator patience.

Third, can Nokia execute on both fronts at once? It is simultaneously pushing a GPU-heavy AI-RAN platform, an anyRAN software layer, a Cohere partnership, and now an open interface strategy. That is a lot of surface area for a company that is trying to revive its 5G position. Focus is a strategy, and Nokia’s current strategy is deliberately broad.

The closing take: openness is the only durable moat left

Here is my argument, stated plainly. The AI-RAN capacity race will be won or lost on compute economics, and compute economics commoditize. Nvidia will sell Aerial-capable GPUs to anyone. Cohere will partner with more than one vendor. The spectral-efficiency gains will converge across vendors within a product cycle or two. When that happens, the only durable moat left in the RAN business is the cost of switching—and the vendor that lowers switching costs for the operator is the vendor that wins the next upgrade.

That is why Nokia’s open interface move is more important than its AI-RAN headlines. It is a bet that the RAN market is finally becoming a software market, where integration and migration matter more than proprietary silicon. Ericsson’s demurral is a bet that it isn’t—that operators will pay a premium for a tightly integrated stack because integration risk is the real cost.

Both bets are defensible. But only one of them is consistent with the direction of every other layer of the network. The core went cloud-native and open. The transport layer went open and disaggregated. The edge is going open. The RAN has been the holdout, protected by physics, real-time constraints, and incumbent lock-in. If Nokia’s new interface is genuinely open—and if the Cohere dApp model delivers what xApps did not—then 2026 will be remembered as the year the RAN finally joined the rest of the network.

If it isn’t, then we will get another five years of open RAN press releases and another five years of flat RAN economics. Operators should watch the governance of this interface, not the spectral-efficiency slide. That is where the answer lives.

Sources

Related Posts

AI-RAN's Commercial Call: 5 Things the Nokia-NVIDIA GTC Berlin Wave Means for Open RAN Operators

Nokia and NVIDIA now count nearly 20 operators on AI-RAN. Five takeaways from T-Mobile's live call, Elisa's field trial and the 2027 pipeline.

AI-RAN Leaves the Lab: A Skeptical Exec's Q&A on T-Mobile's Live Call, Elisa's Field Trial, and Nokia-NVIDIA's 2027 Playbook

A skeptical exec's Q&A on T-Mobile's live AI-RAN call, Elisa's European field trial, and what Nokia and NVIDIA's 2027 commercial timeline really means.