How tightly linked are AI RAN and open RAN?How tightly linked are AI RAN and open RAN?

As open RAN moves toward the adoption and productivity phase, AI RAN is taking over as the exciting new "buzzword" technology with great potential for innovation. Is this a natural transition in which open RAN enables AI RAN, or is it something more disruptive? A question in the new Heavy Reading 2024 Open RAN Operator Survey seeks to better understand how closely linked these two RAN buzzwords are in the minds of operators today.

This is not an easy question to ask in a survey. Since open RAN has been in development, there have been big advances in AI/ML technologies. AI/ML-derived algorithms are now routinely applied in mobile networks and devices, and initiatives are underway — for example, in 3GPP and in the AI RAN Alliance — to use AI more deeply in the RAN.

Yet, at the same time, AI RAN is immature and hard to describe. We have a concept, the outlines of how it might work and a lot of R&D underway, but not much in the way of detail, consensus or specification.

With that caveat in place, the survey asked respondents to take a five-year view of the relationship between AI RAN and open RAN. This is a long enough period for some of today's AI innovations to be productized and deployed in commercial RANs, but also far enough out that responses should be considered speculative.  

The results (see figure below) show a strong link between AI RAN and open RAN — which is predictable. But they also identify a significant constituency that does not necessarily see the two technologies on a straight-line continuum.

Over the next five years, how closely linked are AI RAN and open RAN at your company?

Of the operator respondent base, close to a third (30%) say the two are "very close – we need open RAN for AI RAN." This figure rises to 47% when filtered for those respondents who expect to be in live open RAN deployment by the end of 2025. In other words, the pioneering group of open RAN respondents identified in the survey are more likely to see a closer link to AI RAN than those more conservative on open RAN deployment.

However, the response of the 36% selecting "closely – open RAN helps, but we also need AI in the traditional RAN" is just as important. Given that traditional RAN represents the vast majority of the world's installed radio base stations, it makes sense that operators would also want to apply AI benefits to those assets. Still, this group sees a close link between AI RAN and open RAN.

Just under a third of respondents (29%) see less of a link and believe "AI in the RAN is independent and will be used in traditional & open RAN." This is an important constituency to understand. These respondents are not saying that AI for RAN has no or little value; rather, they are saying that it is not inherently linked to open RAN. Many operators are committed to evolving vendor-integrated RAN over the longer term. Clearly, in these cases, AI could be important over a five-year period.

The outcome is that an overwhelming 95% of respondents are interested in the potential of AI RAN. Beyond that, how people answer this question may come down to definitions of what counts as open RAN and what counts as AI RAN. If you anticipate AI and RAN workloads will run on the same software-defined infrastructure, perhaps simultaneously or scaled according to the prevailing workload demand, then open RAN and cloud RAN are prerequisites for AI RAN.

At the same time, Heavy Reading is aware of companies developing AI-optimized RAN silicon for radio and baseband applications. The idea is to take advantage of AI computation at lower power consumption. This is more of a hardware-centric view of AI RAN, but it may still be considered open in terms of its programmability and potential to be integrated with cloud infrastructure.

There is no question that AI is an important and interesting evolution of RAN. But keep in mind that a five-year view of an undefined technology in rapid development is necessarily speculative.

The full report is available to download here.

This blog is sponsored by Wind River.