AI-native networks will long precede 6G, 80% of telecom operators tell Nvidia

A new report revealed almost 80% of telecom operators expect to see AI-native networks get the jump on 6G deployment, while around two-thirds said AI is driving autonomous network build out.

This is according to Nvidia’s latest “State of AI in Telecommunications” report, which surveyed more than 1,000 respondents worldwide, spanning operators, network equipment providers, system integrators, software vendors, and other players in the telecom ecosystem.

Kanika Atri, senior director for telco marketing at Nvidia, noticed a stark increase in the rate of AI adoption compared to previous years.

“It just blows my mind,” she told SDxCentral, explaining that in many countries, particularly those in Asia-Pacific, agentic AI had already moved from investment into adoption, with telcos now entering an adaptation phase where processes and architectures are redesigned around AI.

“This time the numbers just blow off,” Atri added. “Like 90% people are already saying they saw impact to revenue, more than a third of them saying they saw more than 10% impact to revenue, 90% of them say they saw driving down of costs. So the question of quantifying ROI is a thing of the past. The telco industry is full steam ahead, investing, adopting, figuring out how to use it.”

Atri framed AI adoption as a “no brainer” for telecom operators, arguing that every country and enterprise will build and use it, with telecom “already ahead of the curve.”

Underpinning AI efforts from Nvidia’s perspective are generative models that sit at the bottom of the stack and “apply [themselves] to different categories of telcos’ operations,” with models initially used for operations support systems (OSS) and business support systems (BSS) purposes.

“But networks is the biggest piece now,” Atri said. “Customer service and experience, which is very big … and things like their own operations: IT, HR, and so on. … So generative AI and generative models at the bottom is a core foundation for all the above.”

Of these cohorts, the director said autonomous networks have become the standout AI use case for investment and impact, with agents being deployed on daily, repetitive tasks and complex activities that are highly error‑prone, such as preparing for traffic spikes during major events.

Atri described this as “the most exciting pace of innovation we have seen,” adding that it is now “the No. 1 category when it comes to investments, to ROI, and future plans.”

This is as AI and agents lend themselves “beautifully” to solve the scale and complexity of network data. But while they are looking ahead to fully autonomous networks, telecom operators are still being pragmatic, prioritizing the “80% of the burning problems” they see every day, according to Atri, over the other 20% of extremely complex edge cases.

Atri also pointed to how quickly telecom operators expect to move to AI‑native network architectures such as AI radio access networks (RAN) and distributed environments.

“What was surprising to me was how forward leaning the industry is on the move to the distributed edge, as well as anticipation of AI-RAN architectures ahead of the 6G cycle,” Atri said. “It speaks to the benefits of AI-RAN as an architecture where it’s included in radio signal processing, direct impact on spectral efficiency, energy efficiency. These are burning problems for telcos today. They just cannot have enough spectrum. They can just never have enough energy and AI, and the RAN network is contributing to the bulk of all of this consumption. And here AI is impacting. … This is about optimization and utilization and eventually opening new business opportunities.”

Governance, sovereignty, and regulation are baked into that challenge. When asked whether operators need models and mechanisms to ensure compliance, Atri answered “100%,” stressing the importance of data sovereignty and privacy in the AI discussion.

Nvidia’s report highlighted data-related issues were the biggest challenge for more than half (54%) of operators, up 34-percentage points from 2024. These challenges concern privacy and sovereignty, as well as data silos and data size and complexity.

At an agentic AI panel discussion hosted by Ericsson last year, Girish Mahajan, senior leader for mobile AI data/automation at British operator BT, asked for much closer involvement from hyperscalers on telecom security regulation so that compliant architectures are “baked in” rather than left for each operator to figure out alone.

“This is what telco expects. If you come up with that kind of architecture, then it will make our lives much easier,” Mahajan told an audience.

Atri concurred, stressing that data sovereignty and privacy were “extremely important” for operators and that guardrails, security, and governance must be included in agentic pipelines.

“Those are not easy. Regulations vary from country to country, and this is where 90% of time is spent just fixing that data pipeline,” Atri said. “So yes, that is definitely something that every telco needs to work on.”