Nokia has staked its latest AI-RAN pitch on dApps, and The Mobile Network argues operators should look there first. The Finnish vendor's newest radio access architecture treats the base station as a computing platform, with distributed applications doing the work a central controller once handled.

The Mobile Network's headline — "keep your eye on the dApps" — frames the apps layer, not the silicon, as the decisive piece of the next RAN cycle. That is a pointed read of where vendor lock-in will sit once AI-RAN stops being a slide deck.

What is AI-RAN, and why now?

AI-RAN has hardened from a slogan into a tender requirement over the past eighteen months. Every major RAN vendor now markets one. Ericsson pushes intent-driven automation, Samsung hawks AI-tuned beamforming, and specialists promise double-digit power savings.

The sales pitch usually opens with three numbers: lower energy, higher throughput per cell, and faster fault detection. The harder question — and the one Nokia's dApps framing tries to answer — is what software actually lives on the RAN after the AI bits arrive.

Why do dApps matter here?

A dApp, in this context, is a distributed application that runs close to the radio unit, not in a central core. Think of it as the RAN's version of an app store, where operators and third parties write software that runs on vendor hardware at the cell site.

The economics rhyme with smartphones after 2008. Whoever owns the application layer captures developer attention, which captures roadmaps. Nokia's bet is that an early dApps catalogue beats a faster modem, and that being first to a credible developer story is worth more than another antenna gain.

What has Nokia actually shipped?

Nokia's anyRAN programme, broadened through its GPU partnership with NVIDIA on Layer 1 processing, gives the dApps story a hardware spine. The vendor claims the architecture runs on both x86 and custom silicon, the kind of multi-source flexibility operators now demand after the supply-chain shocks of the early 2020s.

The Mobile Network's framing is also a warning. dApps without a developer ecosystem is just proprietary software with a fresh label. And proprietary RAN software, as O-RAN advocates will remind anyone within earshot, is what the industry spent fifteen years trying to dismantle.

Where do the standards stand?

• O-RAN Alliance runs an active dApps working thread, with no ratified interface yet.
• 3GPP Release 18 mentions AI in the RAN mostly in study items.
• Release 19 remains in early drafting.
• ETSI has started a related workstream on AI management, but no binding spec exists.

Until a body signs off on a common dApps API, every vendor claim is, in legal terms, a private interface. That isn't fatal. Nokia's earlier anyRAN moves did land inside O-RAN's open fronthaul work eventually. The timeline, though, runs in years, not quarters, and operators refresh on cycles their CFOs set, not the ones vendors prefer.

What should operators ask vendors?

• Which dApps ship on day one, and are they Nokia's or a partner's?
• Can the dApps run on non-Nokia hardware, or are you buying a closed stack?
• Who owns the data the dApps generate, and can you export it?

The first tells you whether the platform exists. The second tells you whether you can multi-source. The third tells you whether you're buying software or signing a leash.

Where this leaves the market

AI-RAN capex remains a rounding error on most operator balance sheets, under five percent of RAN spend in the public 2025 roadmaps released so far. MWC25 in Barcelona will set the tone for whether that figure doubles by 2026 or holds flat while marketing outruns engineering.

Nokia's dApps play is the right fight to pick. Whether the vendor has the developer pull to win it is the question The Mobile Network has put on the table, and the one operators will quietly answer at their next refresh cycle — provided their CFOs share the patience.