NVIDIA: AI delivering ROI for telcos, advancing into production

Nearly all telcos are assessing or deploying AI, according to NVIDIA’s State of AI in Telecommunications survey.
More than three-quarters of respondents already seeing return on investment through cost savings and/or increased revenue.
37% of operators focusing investment on applying AI or GenAI in network planning and operations, including integration in radio access networks.
Findings appear to bode well for nascent AI-RAN developments.

NVIDIA: AI delivering ROI for telcos, advancing into production

Source: NVIDIA

Telcos are seeing increasing cost savings and revenue boosts from investments in artificial intelligence (AI) and showing greater willingness to take the technology deeper into their networks, according to new research from NVIDIA, which TelcoTitans discussed with the vendor’s Global Head of Business Development for Telco, Chris Penrose.

The third annual State of AI in Telecommunications report takes the pulse of telcos’ plans and experiences with the technology and the results look encouraging for its author, AI system and software behemoth NVIDIA — with headline finding being that nearly all (97%) telecom operator respondents are assessing or adopting AI.

The technology is already considered to be delivering results for most, as 83% of operators said AI is helping to increase revenues, up from 72% in last year’s survey. Some (21%) have seen AI contribute revenue uplift of more than 10%, while 38% have tracked increases of less than 5%.

AI is also helping operators to reduce costs, with 77% agreeing that the tech has supported decreases in annual operating costs.

“ More than three quarters of companies indicated that they are seeing a positive impact on return on investment [ROI]… [they] are seeing real dollars begin to happen” –Penrose.

Chris Penrose

Source: NVIDIA

Part of the higher ROI realisation is driven by AI projects progressing from proof of concepts to “full production”. As solutions scale out, “that’s really where the big ROI is going to come”, Penrose said.

Telcos are seeing the biggest impacts from AI in higher employee productivity, enhanced customer experience, and improved network operations, the survey found.

AI buds in the network

Improving customers experience remains the top AI use case and investment area for telcos, according to the survey.

The next highest-ranked use cases were network planning and operations (which includes integrating AI into RAN infrastructure) and optimising field operations.

“ The network side has really evolved, with now 37% saying that they’re applying AI and generative AI in the network space, particularly network operations. This is the core of what they do… [They can] apply [AI] in the network itself to drive better performance that could be everything from energy savings to spectral efficiency” – Penrose.

Along with the opportunity to improve how the network performs, Penrose said telcos can have, “for the first time”, a common set of infrastructure that can support the radio access network (RAN) as well as other AI workloads.

“ We’re seeing a lot of interest around this because telcos spend so much capital on proprietary networks today but they’re massively underutilized. They’re designed for the peak hour, but at midnight, you’ve got a lot of spare capacity. We can redeploy those compute assets to do AI and GenAI. It’s a whole new opportunity to create revenue. ” – Penrose.

NVIDIA is one of the founding members of the AI-RAN Alliance, which launched at February 2024’s Mobile World Congress (MWC), along with Amazon Web Services, Arm, DeepSig, Ericsson, Microsoft, Nokia, Northeastern University, Samsung Electronics, SoftBank, and T-Mobile US (Deutsche Telekomwatch, #133). The organisation has since increased to 50 members, including telcos Boost Mobile Network, Globe Telecom, Korea Telecom, and SK Telecom. Demonstrations of the Alliance’s work are expected to be on show at Arm’s stand at MWC 2025. The group’s aim is to: use AI to improve RAN spectral efficiency; integrate AI and RAN to use infrastructure more efficiently and open revenue opportunities; and deploy AI services at the network edge through the RAN. SoftBank is possibly the most advanced operator in exploring AI-RAN’s potential.

In September 2024, T-Mobile US announced a collaboration with Ericsson, Nokia, and NVIDIA to establish an AI-RAN Innovation Centre to study bringing RAN and AI closer together through NVIDIA’s AI Aerial platform.

Telcos ramp up GenAI plans

This year’s survey included a section dedicated to telcos’ plans for, and adoption of, GenAI, finding that the sector is one of the fastest to implement the technology, according to Penrose.

But the “shocking” result for Penrose was that 84% of respondents said they intend to offer GenAI services to their customers. “That’s a really powerful statistic”, he said.

To enable these services, NVIDIA is seeing interest in Europe and Asia for “sovereign AI”, whereby telcos are supporting governments in establishing national AI infrastructure, said Penrose.

NVIDIA has partnered with 14 telcos on sovereign AI solutions, including Iliad Group subsidiary Scaleway, Swisscom and its Italian subsidiary Fastweb, and Telenor.

The national infrastructure can offer anything from AI training and inferencing as a service to GPU-as-a-service for startups and businesses, he explained.

“ This really does provide a new opportunity to drive revenue now but also use the same infrastructure to drive their own cost savings and support the network for the future. And when all those things come together, it becomes a very exciting place for the telcos to lean into. ” – Penrose.

Recent examples of telcos bundling GenAI into their consumer services include Telefónica and Deutsche Telekom offering Perplexity’s digital search assistant. Both operators are also investors in the GenAI scaleup via their respective corporate venture arms, along with NVIDIA. Meanwhile, Orange Business launched a suite of GenAI services for enterprises earlier this year under its Live Intelligence product range. French rivals Iliad and Bouygues offer twelve months free access to AI digital assistants from Mistral AI and Perplexity, respectively. In the UK, EE’s 5G standalone network launch in September last year was pitched as part of the operator’s efforts to prepare for an expected influx of AI-powered devices.

Finding AI talent remains top challenge

The survey also found some persistent challenges to telco adoption of AI, with the top three obstacles the same as in the 2024 survey.

The biggest hurdle to scaling AI is lack of talent, such as finding data scientists, engineers, architects, and developers.

More respondents this year (38%) pointed to ’inability to quantify ROI’ as an issue, up from 33% last year, while 30% cited lack of budget for AI as problematic.

**Key findings from NVIDIA’s State of AI in Telecommunications survey**

84% — AI is helping to increase their company’s annual revenue.
77% — AI helped reduce annual operating costs.
97% — Either actively deploying or assessing AI projects.
60% — Increased employee productivity the biggest benefit from AI.
44% — Investing in AI for customer experience optimisation, which is the top area of investment for AI in telecommunications.
40% — Deploying AI into their network planning and operations, including RAN.
49% — Adopted GenAI solutions.
84% — Using GenAI plan to offer solutions to the customer.