Telcos are ‘aggressively’ leaning into AI, Nvidia report reveals

Of the telecom professionals surveyed by Nvidia, 97% said their company is either actively deploying or assessing deploying AI

A new survey published this week by Nvidia reveals that telcos are leaning into artificial intelligence (AI) , and even generative AI (gen AI), quite aggressively, going against the industry’s reputation for being slower to adopt new technologies. In fact, according to NVIDIA’s Global Head of Business Development for Telco Chris Penrose, telcos, by many accounts, appear to be leading in the adoption of these technologies.

“And part of the reason why is the opportunity,” he told RCR Wireless News. “These are large companies with lots of legacy systems and lots of employees and lots of places where AI — and generative AI in particular — can really help transform the cost structure, transform the customer experience, make their employees more productive and ultimately, drop dollars to the bottom line and help them make more revenue… telcos have leaned into AI more aggressively … [and] we are starting to see some real results behind that.”

Of the telecommunications professionals surveyed by Nvidia, 97% stated that their company is either actively deploying or assessing deploying AI in their operations, with 80% of respondents saying they believe AI to be crucial for their company’s future success, and two-thirds planning to increase spending on AI infrastructure this year. Top areas of AI investment include improving network planning and operations (37%) and in using AI for field-operations optimization (33%).

Additionally, 58% said increased employee productivity was their biggest benefit from AI, up from 33% last year, followed by enhancing customer experiences (37%) and improved network operations (32%).

And perhaps most importantly, 84% said AI is already helping to increase their company’s annual revenue and 77% said it helped reduce annual operating costs.

Generative AI

Nvidia has conducted a telco AI survey for the past three years, but this is the first time it has included questions around gen AI: “When we started the report, generative AI wasn’t even a term,” explained Penrose, adding that this year, its inclusion was a priority. “What has emerged in this last round is that telcos are using generative AI in the network itself, and this idea around employee productivity and many telcos taking the corpus of information that they have and making it a lot easier to interface for their employees – it could a be network technician in the fields, it could be a care agent, but it could also even be people that are working in like the legal department, supply chain and so on,” he continued.

Specifically, the survey showed that of those respondents adopting generative AI, 84% said that their companies plan to offer generative AI solutions externally to customers. 52% said they would offer generative AI as a software-as-a-service solution, while 35% will offer generative AI as a platform for developers, including for compute services.

The survey also revealed a shift from closed-source AI models towards open-source model, said Penrose. He said that “spike” in how much telcos want to tackle AI themselves tells him that they are realizing that their data is “gold” and that if they can apply AI and gen AI, they can create a better experience, a better product and even new revenue opportunities.

Collocating

Another emerging opportunity captured in the Nvidia report is around collocating AI applications and RAN applications on common infrastructure, a concept that the AI-RAN Alliance calls AI-RAN. While not the most popular response to how they will invest in AI for 5G monetization and/or 6G research and development (that goes to deploy AI services on RAN for operational and user needs at 66%, followed by enhancing spectral efficiency for the RAN at 53%), 50% did say they are planning to collocate AI and RAN applications on the same infrastructure.

And this marks a key shift in the telco approach: “For a long time, it’s been kinda like ‘never — my network is my network,’” said Penrose. “But the fact is, that it means that you have a very underutilized asset… with stranded and wasted compute infrastructure.” That’s because telecom networks are designed for peak and when you design something for peak, that actually means that most of the time, it’s not being used to the fullest extent. By running and managing RAN and AI workloads concurrently, network resources can be dividing up based on time of day or on the amount of compute.

Last November, SoftBank put this approach to the test when it piloted the world’s first combined AI and 5G telecom network using the NVIDIA AI Aerial accelerated computing platform in an outdoor trial conducted in Japan’s Kanagawa prefecture. The companies said that concurrent AI and RAN processing was successfully demonstrated between RAN and AI workloads, with the goal of maximizing capacity utilization.

Nvidia claimed that AI-RAN enables telcos to achieve almost 100% utilization compared to 33% capacity utilization for typical RAN-only networks — an increase of up to 3x — while implementing dynamic orchestration and prioritization policies to accommodate peak RAN loads. The pair also stated that this capability transforms base stations from “cost centers into AI revenue-producing assets,” with the pair estimating that telco operators can earn approximately $5 in AI inference revenue from every $1 of capex it invests in new AI-RAN infrastructure, achieving a return of up to 219% for every AI-RAN server it adds.