Nvidia’s 6G bet: How AI will shape the next-gen wireless network

  • Nvidia is working with T-Mobile, Cisco and others to build an AI framework for 6G

  • Analysts had lots to say on the planned 6G network stack

  • Will AI dominate 6G standards?

Nvidia is trying to get an AI head start with 6G.

As Fierce reported this week, Nvidia is collaborating with T-Mobile, MITRE, Cisco, ODC and Booz Allen Hamilton to develop an AI-native network stack for 6G. The companies are basing this work on Nvidia’s AI Aerial platform. This follows on from Nvidia’s launch of a 6G research cloud platform at the GTC conference in 2024.

Despite the massive amounts of hype around 6G, it will still be at least four and a half years before the forthcoming standard actually goes commercial. 6G isn’t even expected to be standardized until 2028 with 3GPP’s release 21. 

Why is Nvidia pushing AI for 6G?

So Fierce asked analysts what Nvidia is trying to do so far ahead of arrival of the commercial standard.

“Nvidia is trying to get ahead of the curve to insert their technology into the 6G standards,” said Mobile Experts analyst Joe madden in an email to Fierce. “That’s how the game is played.”

“What NVIDIA is doing with T-Mobile, Cisco, and other companies currently is working to better understand how to make AI native to the 6G standard,” noted Daryl Schoolar, analyst at Recon analytics. “Answering such questions as what it specifically needs to do, and how to make it part of the standards. By answering those questions NVIDIA and its AI RAN partners hope to influence what goes into those standards.”

This is to try and get 6G infrastructure built around the AI chips that Nvidia favors, its graphical processing units (GPUs) rather than the x86 chips and custom ASICs that operators favor today. “NVIDIA is hoping that they can get  traction in the research community, along with some ecosystem support and perhaps even some non-incumbents (doesn't have to be a small startup, could be a networking vendor with mobile core but no mobile RAN assets like say Cisco or other) to bet on the next S-curve leveraging GPUs instead of the traditional ASIC route,” said Roy Chua, lead analyst at AvidThink.

Where are the major RAN vendors?

The latest Nvidia 6G collaboration doesn’t include any of the major RAN vendors. But as Chua points out the company is already working with Ericsson, as well as Nokia. It may, however, take until the advent of 6G next decade to tame the costs, complexity and power-hungry nature of AI RAN. Intel’s CTO recently told Fierce that AI RAN doesn’t make sense cost-wise today.

Still, AI-native 6G is an evolving battleground. Madden noted that he doesn’t think that Nvidia’s way is the only way to create an AI native wireless network. “Nvidia has very little support from operators and vendors in the RAN community, so it’s unlikely that their particular flavor of AI hardware implementation will be considered the foundation for 6G,” Madden claimed.

We will see how the push to define and standardize AI native in 6G will develop over the next couple of years. Doubtless, Nvidia, Intel, Qualcomm and others will have plenty to say on how AI-native wireless networks should advance for 6G.