• NVIDIA and SoftBank announced their plans to transform their 5G network into a revenue generator by deploying AI-RAN at AI Summit Japan 2024.

• The move signals NVIDIA’s intentions to diversify its business amid questions about the sustainability of hyperscaler spending.

• This announcement is one of several initiatives by NVIDIA across the sector to help telcos build next-generation services using LLMs for the government, enterprise and consumer sectors.

• Back-of-the-envelope estimates suggest total addressable market for NVIDIA could reach US$200bn globally in an aggressive deployment scenario.

• These moves could indicate NVIDIA plans to take a larger role in the telecom vendor ecosystem.

NVIDIA CEO Jensen Huang recently visited Japan as part of a global tour promoting the future development of AI ecosystems in several key economies. The highlight of the company’s first “AI Japan Summit” was a conversation between Huang and SoftBank’s Masayoshi Son about the future of 5G networks and the pivotal role AI will play in transforming mobile networks.

The conversation was notable in that it was clear that the mobile network space is an increasing focus for the company. And it was equally clear that mobile network operators are exploring radical new approaches to network management going forward.

Both companies have made bold claims about the earning potential of this future AI-RAN network. They claim to be able to generate US$5 in AI inference revenue from every US$1 of capex invested in AI-RAN infrastructure over the lifetime of this new equipment.

It makes sense for NVIDIA to focus on the mobile network space as because of the sheer market opportunity: There are roughly 10 million 5G base stations in the world now and adding intelligence with an NVIDIA chip could easily translate into billions of dollars in revenues for the company.

The Multi-Billion Dollar AI-RAN Opportunity – For Telcos & Vendors

According to Japanese government data, SoftBank had 65,366 5G base stations in Japan at the start of 2024. Assuming that an AI-RAN server can service 20 cell sites and an NVIDIA GH200 Grace Hopper Superchip server costs roughly US$400,000, this equates to a US$1.3bn investment from SoftBank to upgrade the entire network.

And on the revenue side, multiplying this figure by five times would equate to a US$6.5bn dollar opportunity for SoftBank. If every 5G network in the world upgraded to AI-RAN and the 5x metric were true around the world this would equate to a roughly US$200bn opportunity for vendors and a US$1 trillion-dollar opportunity for telcos. Clearly this is a best case, back of the envelope calculation, but it should come as no surprise that this topic is attracting much attention recently.

It is more likely that SoftBank and other telcos would initially upgrade the network in dense urban areas where roughly half their 5G network is located – which would still equate to a US$650 million investment with US$3.3bn in additional revenue for SoftBank with the global opportunity standing at US$100bn for vendors and US$500bn for telcos; a game changer for the industry.

Back of the Envelope: Global AI-RAN Vendor Opportunity, Urban & Total Deployments (US$bn) TAM

Source: Counterpoint Research

From SoftBank’s perspective introducing AI into their mobile network also makes sense. The telco’s ARPU has not increased since it launched commercial 5G services in 2020. In fact, its ARPU, like the industry average, continues to fall despite all the additional CAPEX and operating costs associated with running a 5G network. This is why telcos are increasingly looking to AI to both optimize network costs by automating as many functions as possible and to create new services. Hence, we expect announcements like the one made by NVIDIA and SoftBank to become more common in the future.

Access full 4-pager report here.

If you are not a client, reach out to us at [email protected]