LIVE FROM UNWRAPPED: THE 5G EVOLUTION: Soma Velayutham, VP AI and Telecoms at Nvidia, argued there is a huge misconception around the price of its GPUs and a gross misunderstanding of the AI powerhouse’s platform, as he opened the door to striking similar vendor deals following its recent €1 billion tie-up with Nokia.

Speaking to Mobile World Live on the opening day of Unwrapped, Velayutham hit back at suggestions operators were not convinced by the economics of introducing GPUs at cell sites, stating it was actually the “best price performing platform” available.

“You get a much better return on investment in GPU,” he said.

He explained most people associate Nvidia with large data centre investment, pointing to companies like OpenAI, Microsoft and AWS buying up its Grace Blackwell platform for this very purpose.

However, he added: “They actually may not know GPUs exist in robots, in cameras, in laptops. We have GPUs that scale all the way down to 3 watts… and can scale all the way up to 1,000 watts in a data centre.”

By utilising Nvidia’s application platform Cuda, Velayutham explained a company like Nokia could now essentially write its software on a 3 watt GPU or a 3,000 watt GPU “or anything in between”.

“It’s a myth that Nvidia GPUS are expensive or power hungry. It’s actually a gross misunderstanding of our platform.”

AI-RAN, 6G
Staying with Nokia, Velayutham backed his CEO Jensen Huang’s claims that the deal would ignite the AI-RAN revolution, while pointing to wider momentum around this area with the AI-RAN Alliance.

He said the company was looking forward to working with Nokia to “deliver mass adoption of AI-RAN”, and while the Finnish vendor was the first to lead the way, the path was clear for other similar partnerships. “We would love to have all vendors – Ericsson, Samsung, Fujitsu, NEC – all of them adopt our platform.”

In terms of an AI-RAN roadmap, the Nvidia executive said there were multiple approaches at play around infrastructure as well as adoption in niche markets like enterprise, defence and the public sector.

For MNOs, he believes live trials will be conducted in the field by the end of 2026, before mass adoption towards the end of 2027.

Velayutham also touched on the impact of AI RAN on future 6G networks, backing up comments made by operator Orange among others that the next-generation will be more akin to a software upgrade rather than a complete overhaul.

“Moving from 5G to 6G is going to be a continuum. Its going to be a software journey not a hardware journey,” he concluded.