Nokia and SoftBank Aim AI-RAN Trial at Commercial Traffic in 2027

Nokia and SoftBank Corp will run an AI-RAN field trial carrying commercial traffic in Japan in 2027, the Finnish vendor announced this week. The date itself is the story: AI-RAN has spent years as a slide-deck promise, and this puts a real deadline on it.

The trial moves the technology from laboratory development into a live network environment. Nokia frames it as a milestone in its own AI-RAN commercialisation roadmap, not just a favour to a friendly operator.

What will actually run on the network?

Nokia will combine its AI-native anyRAN cloud RAN software with SoftBank's core AI-RAN platform, AITRAS, plus Nvidia's ARC platform. The pitch is straightforward: run radio workloads and AI workloads on one common platform instead of two separate estates. That, Nokia argues, lets operators "improve infrastructure utilisation, accelerate innovation and unlock new revenue opportunities."

Note the wording. Utilisation is measurable. "Unlock new revenue opportunities" is vendor vocabulary for revenue that doesn't exist yet. The two claims deserve separate scrutiny, and the 2027 trial should help distinguish them.

Who said what

SoftBank SVP and CTO Ryuji Wakikawa called the trial a significant milestone in moving AI-RAN from concept to execution. "We are excited to begin this next phase of validation with Nokia, paving the way for a new era of networks that can deliver both connectivity and AI services," Wakikawa said.

Pallavi Mahajan, chief technology and AI officer at Nokia, pointed to the technology's dual promise. AI-RAN can "unlock new levels of spectral efficiency while providing the distributed intelligence needed to support physical AI," Mahajaan said.

Spectral efficiency is the concrete half of that sentence. Physical AI — robots, autonomous systems, machines that reason at the edge — is the speculative half, and it leans on infrastructure nobody has fully built yet.

Why this isn't a cold start

The two companies already operate an AI-RAN agreement, backed by an Nvidia accelerated computing node. The 2027 trial extends an existing relationship rather than launching a new one, which usually means fewer integration surprises.

For SoftBank, the Japanese operator has been among the loudest advocates of AI-RAN as a category. A live-network validation under commercial traffic would give it evidence that most operators interested in the concept still lack.

For Nokia, the stakes are commercial. Cloud RAN and AI-RAN sit at the centre of its effort to stay relevant as operators rethink who builds their radio networks and how. A named operator, a named market and a named year beat a roadmap slide every time.

What happens between now and 2027?

The gap between announcement and trial is roughly two years, which tells you how early AI-RAN still is. The industry consensus has shifted from whether telecom networks and AI compute should converge to how quickly, but live deployments under paying-customer traffic remain rare.

If the 2027 trial delivers measurable gains in spectral efficiency and platform utilisation on SoftBank's network, expect the results to anchor AI-RAN procurement conversations across the industry. If it doesn't, the concept gets another slide and another wait. Wakikawa's "concept to execution" framing sets the test: 2027 is when the slide meets the subscriber.