AI-RAN gathers momentum with Ericsson-Softbank tie-up

The concept of so-called AI-RAN sounds fairly simple, by telco standards at least. Instead of hosting AI applications and virtual RAN (vRAN) software on separate servers in different locations, the aim is to host both of them on the same vRAN infrastructure.

This integration provides an opportunity to enhance RAN operations by analysing KPIs, predicting traffic patterns, and automatically adjusting the network. In addition, it can also serve as a platform for offering hosted AI applications to enterprise customers.

SoftBank claims this is tantalising prospect because telcos can spin up two revenue streams – network access and AI services – from a single capital outlay.

Indeed, SoftBank is very bullish about AI-RAN, claiming it could not only transform the business of telecoms, but also – in its own words – have "a profound effect on society as a whole."

It was one of the 11 founding members of the AI-RAN Alliance, which aims to coordinate an industry-wide effort to develop and promote AI-RAN technology. Launched at this year's Mobile World Congress (MWC), its membership includes Nvidia, Microsoft, Amazon Web Services (AWS), Arm, Samsung, T-Mobile US, and Nokia.

Ericsson is also a member, and under its new partnership with SoftBank it aims to identify common network and compute infrastructure that lends itself well to hosting AI and RAN software alongside one another.

The work will involve joint techno-economic analyses, prototype development, and lab demos to optimise RAN and AI convergence at the edge.

Ericsson and SoftBank will also try to figure out hardware partitioning and workload distribution between AI and RAN – presumably it is important to find the right balance in all this so as to mitigate the risk of degrading the performance of either. They will also look into software portability between different hardware platforms.

"We look forward to this collaboration with SoftBank in exploring the potential convergence of AI and RAN infrastructures. This reaffirms our commitment to innovation and excellence, and we believe it will lead to new solutions that empower communications service providers to build more open, efficient and versatile networks," said Fredrik Jejdling, EVP and head of Ericsson's networks business.

"SoftBank welcomes this new collaboration with Ericsson, which aligns with our strategy to invest in AI infrastructure that enables the overlay and optimisation of RAN. This partnership reflects our vision of leveraging AI to enhance communication networks and opens up opportunities for collaboration with key industry players," added SoftBank EVP and chief technology officer Hideyuki Tsukuda.

This development comes not long after T-Mobile announced it was getting together with Nvidia, Ericsson, and Nokia to open an AI-RAN innovation centre, with the aim of advancing this nascent tech.

Meanwhile, as well as giving telcos a means of enhancing their networks, the AI-RAN movement appears to be very closely aligned with Nvidia's interests.

It has emerged as the chip maker of choice when it comes to AI hardware, but faces stiff competition from rival Intel in the vRAN market. Successfully running AI and the RAN on the same platform is likely to place greater demand on compute resources, potentially necessitating something with a bit more oomph – like an Nvidia graphics processing unit (GPU), for instance.

With Intel struggling financially, the last thing it needs is Nvidia making greater strides into the RAN market. But if AI-RAN continues to gather momentum, in all likelihood that's exactly what will happen.