Interview: Freshwave CTO offers AI-RAN walkthrough

As part of our month-long in-depth look at AI-RAN, we thought it would be interesting to get the perspective of neutral host network providers to find out if current infrastructure is prepared for the technology or would require changes. First up is Freshwave Group, a UK-based company which works with domestic MNOs to provide multi-operator indoor coverage for hundreds of enterprises, hospitals and central government properties, along with holding contracts with more than 40 local authorities to deliver enhanced 4G and 5G connectivity at street level. Its CTO Tom Bennett ( pictured ) is pretty optimistic the overall concept of employing AI in RAN operations is achievable, but said the foundations must be in place to reap the fullest benefits. Data is the key groundwork. Bennett said because the LLMs AI relies on to function are based on vast amounts of data, it is essential to develop ways to share the information to enable the learning required for telecoms purposes. “So, do I think that the network needs a whole bunch of GPUs in it? No, I don’t. For the avoidance of doubt, I really don’t see the benefit of adding a GPU to a bunch of cell sites or small cell sites or DAS sites in hospitals”. Instead, Bennett believes data sharing is “probably the first and key thing”, because it could enable RAN sites to communicate information to an AI located elsewhere. Data centres are an obvious target: you don’t need to look far for headlines about the massive amounts of GPUs being ordered and installed in these facilities. But Bennett noted there needs to be a common framework to get the data from the RAN to the data centre before AI can do its stuff. He believes a common language is required which can transmit key details regarding signal strength and quality, fade levels and the effects of the landscape on the beam. Fibre is Bennett’s preferred method of connecting RAN sites to a data centre. “The speed of light on a fibre is about 1 millisecond per 100kms, so you don’t need the GPU to be right up in the cell side”. Freshwave employs fibre in the “vast majority” of its deployments: Bennett highlighted the required bandwidth as one crucial element the connectivity technology brings to the mix and noted deployments are largely now beyond the minimum needed to maximise use for cloud and distributed RAN purposes. Employing fibre to essentially move AI elements off specific RAN equipment is where Bennett believes the main opportunities for operators lie. MNOs “can opti