5G & 6GAI RanIsacNvidiaEdge Computing
Nvidia targets mass AI-RAN deployment in 2027 as ISAC gains speed
Nvidia expects mass AI-RAN deployment in 2027 after field trials exceeded expectations, with ISAC contracts already signed in public safety, its telecom VP said.
5G & 6GWhy it matters
- Nvidia targets mass AI-RAN deployment in 2027 after 2026 field trials slightly exceeded expectations
- Nvidia holds ISAC field trials and contracts in public safety, two years after the topic had little traction
- Nvidia names voice AI, AI video and robotics as three credible edge use cases
- Nvidia exec said ISAC was expected on the 6G roadmap but is arriving much faster
The story
Nvidia now expects mass AI-RAN deployment in 2027, after operator field trials in 2026 slightly exceeded the company's expectations. Soma Velayutham, VP for AI and Telecoms at Nvidia, delivered the timeline during the second keynote at the Unwrapped – The 5G Evolution event, where he also admitted that traction around integrated sensing and communications (ISAC) has caught the chip vendor off guard.
The admission is notable because Nvidia had placed sensing firmly on the 6G roadmap. "We were thinking ISAC is somewhere down the road of 6G, but it is happening much, much faster," Velayutham said. Two years ago, he noted, there was little industry talk about ISAC at all. That has changed into live commercial activity rather than lab work.
"We've now got field trials, in fact, we have contracts in public safety and other areas," he said. Sensing, he added, "has moved much faster than we thought."
For practitioners, the significance is straightforward. A technology the vendor community had bracketed with 6G timelines is generating signed contracts today, with public safety as the early anchor market.
Which AI-RAN use cases does Nvidia consider credible now?
Beyond sensing, Velayutham named three categories that make commercial sense at the edge today: voice AI, AI video and robotics.
Voice AI stands out for its interactivity requirements. "The pause between the time to token is very important," he said, referring to the latency between a user's speech and the model's first output token. He added: "Being able to deliver a conversational AI that's very fluent is something that's important that kind of demands the right memory architecture placement."
Video earns its place on economics, not latency. Velayutham argued video is credible at the edge "not necessarily because video is latency sensitive", but because the sheer volume of data involved costs a lot of money and time to move in and out of the cloud.
Robotics is the newest and least mature of the three. Humanoids and mobile robots face two hard constraints: a memory threshold and battery limitations. Edge compute offers a workaround that shifts the burden off the device itself.
"One of the things we are seeing is how edge compute can lease the compute or rent the compute to the robot and the robot basically is using compute at the edge and therefore reducing its own energy consumption and being able to extend its useful life," Velayutham said.
The model is effectively compute-as-a-service for machines. The robot rents processing capacity from nearby edge infrastructure instead of carrying it onboard, cutting energy draw and extending operating life between charges.
What else is on the agenda?
Velayutham also used the interview to discuss Nvidia's AI-RAN partnership with Nokia, one of the vendor pairings formed to bring GPU-accelerated computing into radio access networks, and to address what he framed as misconceptions around the cost of Nvidia GPUs.
The cost question matters for operators weighing AI-RAN business cases, since GPU capex is the most common objection raised against the architecture.
Nvidia's stated sequence now runs from trials in 2026 to what Velayutham called "mass deployment" in 2027, with sensing contracts already signed in public safety and edge workloads in voice, video and robotics forming the near-term commercial base.
Also reported
Source: Mobile World Live
More from Daniel Okafor
Show full bio
Senior reporter covering regulation, policy and their impact on Carrier business, 5G and 6G, fiber and broadband at Telecom Gazette.
559 articles