Nokia’s dApp Bet: The E3 Interface and the Fight for Open RAN’s Next App Economy

If you followed the news out of NVIDIA’s GTC Berlin event this week, your feed was probably full of one story: T-Mobile, Nokia and NVIDIA completing what they called the first AI-RAN call on a live commercial 5G network in Bellevue, Washington. That’s a genuinely significant milestone, and it deserves the attention it got. Nearly 20 operators worldwide have now announced AI-RAN collaborations built on Nokia’s anyRAN software and NVIDIA’s Aerial RAN Computer, according to the companies.

But buried in the same wave of announcements was something quieter, and possibly more consequential for anyone who cares about Open RAN as an open, multi-vendor ecosystem. Nokia said it plans to offer so-called dApp software plug-ins through the Open RAN standardised E3 interface, as part of its AI-RAN ecosystem. That’s a mouthful of jargon, so let’s unpack it — because it points at a question the industry has been dancing around for years: once networks get smart, who gets to write the apps that make them smart, and who collects the revenue when they do?

First, what on earth is a dApp?

Most people hear “dApp” and think of blockchain. In the telecom world, it means something different: a distributed app, a small software module that runs inside the radio access network — the part of a mobile network that connects your phone to the core — and does something useful in near-real time.

To understand why this is a big deal, you need a little context on how Open RAN has evolved. The original promise of Open RAN was disaggregation: break the radio access network into pieces from different vendors that all speak common languages, so operators aren’t locked into a single supplier. Alongside that came a software layer of applications that run on top of the network’s controllers. The first wave were called rApps — non-real-time applications that run on the RAN Intelligent Controller, or RIC, and handle things like policy, optimisation and longer-horizon automation. Think of them as the network’s back-office brain.

But some decisions can’t wait. If a radio needs to adjust how it handles a signal in milliseconds, an rApp is too slow. That’s where dApps come in: near-real-time applications that sit closer to the action, inside the distributed unit or the radio unit itself. And the interface that lets them plug in and talk to the network in a standardised way is called E3.

If that sounds like plumbing, that’s because it is. But plumbing is exactly where ecosystems get built or strangled.

Why this announcement matters more than it looks

Here’s the thing about interfaces: whoever defines them, and whoever builds the first real marketplace on top of them, tends to capture the value. Apple didn’t win the smartphone era because it made the best phone. It won because it made the best place for other people to sell things.

The same logic applies to Open RAN. For years, the industry has talked about an “app economy” for the RAN — a world where third-party developers write software that runs on any operator’s network, and operators pick and choose the modules that improve coverage, save energy or boost capacity. That vision has been mostly theoretical, partly because the standards work on interfaces like E3 has been slow, and partly because the big vendors haven’t had much incentive to open the door.

Nokia’s move is notable because it’s a major incumbent vendor saying, in effect: we’ll support the standardised plug-in route. According to Telecompaper, Nokia said it would offer dApp software plug-ins as part of its AI-RAN ecosystem via the Open RAN standardised E3 interface. The report also notes that this follows planned product support for dApps by US vendor Cohere Technologies.

That last detail is important. Cohere has been pushing dApps for a while, and if you’ve been following this blog, you’ll know we covered Cohere’s E3 work as a potential unlock for scalable AI-RAN applications. Now a tier-one vendor is signalling it will play in the same sandbox. Two data points don’t make a market, but they do make a direction of travel.

The bigger picture: AI-RAN needs somewhere for the intelligence to live

Step back and look at the week as a whole. The T-Mobile call, the SoftBank trial targeting 2027, the Orange testing, the BT trial with Nokia’s 5G software on NVIDIA’s Aerial RAN Computer, Elisa running voice, mobile data and video analytics simultaneously on the same AI-RAN platform in Helsinki — all of it points in one direction. Networks are becoming compute platforms, and the intelligence running on them needs a home.

That’s where the E3 conversation gets interesting. If AI-RAN is going to work economically, operators need to be able to mix and match AI workloads — some from the vendor that sold them the radio, some from third parties, some built in-house. Without standardised interfaces, every integration is a bespoke project, and bespoke projects don’t scale.

Nokia’s framing at GTC Berlin was about “unlocking more distributed, instantaneous applications” — the kind of thing that only works if the software can plug into the network quickly and reliably. That’s the pitch. Whether it becomes reality depends on how genuinely open the implementation turns out to be, and whether other vendors follow.

The sceptic’s corner

Let’s be honest about the caveats, because there are several.

First, this is a plan, not a shipped product. Nokia said it would offer dApps via E3. The gap between “will support” and “operators are running third-party dApps in production” is usually measured in years, not quarters.

Second, “standardised E3 interface” is doing a lot of work in that sentence. Standards are only as useful as their implementations, and the Open RAN world has plenty of examples where a standard existed on paper but interoperability in the field was another matter entirely. The E3 specification is still maturing, and there’s a real question about how much of the intelligence can actually run in near-real time without hitting latency walls. We’ve written before about why the sub-10-millisecond battleground matters; dApps live or die on that physics.

Third, and most importantly: who controls the marketplace? If Nokia’s dApp ecosystem is genuinely open — any developer can write a module, any operator can deploy it, and the interfaces are truly standard — that’s a big deal. If it’s a curated catalogue where Nokia decides what gets in and takes a cut, it’s a much smaller deal, and arguably a clever way to keep the app economy inside the vendor’s walls while appearing to open it up.

We don’t know which it will be. The announcement didn’t say. That’s not a criticism so much as a flag: watch the implementation, not the press release.

Why operators should care

The business case for dApps, if it works, is straightforward. Operators spend enormous sums on spectrum and radios, and the marginal gains from squeezing more efficiency out of that hardware are where the money is. Nokia has reported spectral-efficiency improvements of more than 20% in its wider AI-RAN platform testing, and Deutsche Telekom’s work with Nokia on AI-based channel estimation reportedly showed typical uplink throughput improvements of 10–15% in initial lab measurements, with greater gains under more complex radio conditions.

Those are the kinds of numbers that get a CFO’s attention. But here’s the catch: those gains came from AI capabilities developed by the vendor. The promise of dApps is that you could get similar gains from a broader pool of innovators — smaller companies, specialist AI firms, even the operator’s own engineers — without waiting for a single vendor’s roadmap.

For an operator, that’s the difference between buying a feature and owning a capability. It’s also the difference between a network that improves at the pace of one company’s release cycle and one that improves at the pace of the entire software industry.

The competitive angle nobody’s talking about

There’s a geopolitical and competitive subtext here too. The AI-RAN story is increasingly a story about who supplies the compute. NVIDIA has invested $1 billion in Nokia, and its Aerial RAN Computer platform is showing up in trial after trial — SoftBank’s 2027 trial will use NVIDIA Grace CPUs, RTX PRO 4500 GPUs and ConnectX-8 networking cards, for instance.

That concentration matters. If AI-RAN becomes the dominant architecture and the compute layer is dominated by one supplier, the “open” in Open RAN starts to look a bit thinner. Standardised interfaces like E3 are one of the few levers that can keep the software layer contestable even as the hardware layer consolidates. In that light, Nokia’s dApp announcement isn’t just a product line extension — it’s a bet that an open app layer is better for the ecosystem than a closed one. Whether that bet is sincere or strategic is for the market to judge.

What to watch next

The next twelve to eighteen months will tell us a lot. Watch for three things.

First, real deployments. SoftBank is targeting 2027 for a commercial AI-RAN trial in Japan, with distributed deployment and evaluation of customer use cases and commercial AI-RAN services expected around then. If dApps show up in that trial with third-party modules, that’s a meaningful signal.

Second, whether other vendors follow. Cohere is already there. If Ericsson, Samsung or others announce E3 dApp support, the interface becomes a genuine standard rather than a Nokia feature. If they don’t, we’ll know the app economy is still more slogan than substance.

Third, the developer question. The rApp market never quite took off the way its advocates hoped, in part because the barrier to entry for developers was high and the customer base was small. If dApps are going to succeed where rApps stumbled, someone needs to make it easy for a small team to build, test and sell a module. That’s a developer-relations problem as much as a technical one, and it’s the part of the story that gets the least attention.

The bottom line

The T-Mobile AI-RAN call was the flashy headline this week, and it earned its coverage. But the Nokia dApp announcement may be the one that shapes what kind of industry we get. A network where intelligence can be plugged in from anywhere is a very different network from one where it comes from a single catalogue. The E3 interface is the hinge on which that choice turns.

Nokia has said the right things. Now the industry needs to see the code, the catalogue and the commercial terms. That’s when we’ll know whether Open RAN’s app economy is finally arriving — or whether it’s just moving into a nicer showroom.

Sources

• Nokia to offer dApp software features via Open RAN E3 interface - Telecompaper
• Nokia, Nvidia accelerate AI-RAN, physical AI commercial deployment - Computer Weekly
• SoftBank, Nokia target 2027 for commercial AI-RAN trial in Japan - Light Reading
• Orange puts AI-RAN through its paces as NVIDIA, Nokia hail momentum - TelcoTitans
• AI Infrastructure Race: 9 Models in 5 Weeks - Tech-Insider
• How First AI Network Call Will Impact T Mobile Stock Investors - Yahoo Finance
• AI-RAN and the Geopolitics of 6G - Debug Lies News