Dell’Oro confirms RAN market hit its steepest decline in 20 years but vendors are hoping AI-RAN will revive the market’s fortunes
Analyst house Dell’Oro has confirmed what people in the industry were already feeling – the RAN market just recorded the steepest full-year decline in more than 20 years. With two consecutive years of steep declines, worldwide RAN is now down nearly $9 billion in 2024 relative to the peak in 2021.
The analysts had the sensitivity to hold off their findings until the close of the world’s biggest mobile festival wound down in Barcelona. But on the corridors of the stands at MWC, talk was already about the fast-emerging next generation RAN, infused/boosting/performance-enhanced by AI. But how fast? Téral Research told SDx recently that the global AI-RAN market will grow from $1.7 billion this year to $10.4 billion in 2030, adding that the “current ecosystem is vibrant.”
ABI Research, on the other hand, reckons in 2025, trials and pilot projects will advance the technology, but large-scale deployment will take longer to materialise. “Despite rapid progress, AI technology for the RAN ecosystem (AI-RAN) remains in early development, with commercial deployments not expected before 2026,” they wrote. “SoftBank recently completed a trial and plans deployment by 2026, with broader adoption likely by 2027, given Japan’s history as an early tech adopter.”
GSMA found that operators are allocating more resources to in-house and out-of-house AI projects, but only a some are spending more than 15% of their digital budgets on AI. Instead, nearly half of operators are dedicating 5% to 15% of their digital budgets towards AI, covering a range of categories, including data systems, large language models and infrastructure upgrades, according to the survey.
At MWC, SoftBank announced that in collaboration with Nvidia and Fujitsu, conducted the three AI-driven experiments to enhance Radio Access Network (RAN) performance without deploying additional base stations. The first experiment, Uplink Channel Interpolation, aimed to improve channel estimation accuracy for signals received at base stations. By applying AI, SoftBank achieved approximately a 20 percent improvement in uplink user throughput on smartphones, particularly in areas with poor network quality.
The second experiment focused on predicting Sounding Reference Signals (SRS) from devices to enhance beamforming performance in high-mobility scenarios. Utilising a Multilayer Perceptron (MLP) algorithm, SoftBank observed around a 13 percent improvement in downlink throughput for devices moving at 80 km/h. The third experiment involved AI-driven Medium Access Control (MAC) scheduling to optimise user pairing in multi-user MIMO scenarios. Implementing AI technology in MAC scheduling led to an approximate 8 percent increase in average user throughput.
What wasn’t discussed as much however was exactly how so many mobile base stations will be able to cope with power-sucking Nvidia GPU chips. Hosting AI inferencing workloads at the network edge simply drives power shortfalls/issues to a widely distributed edge network. Some of this AI inferencing will actually shift to on-device but the edge power situation does not resolve itself as a result. Thankfully, there are some very clever chip companies attacking these issues globally but the net result for telcos is that AI-RAN will tend to develop more slowly, pretty much like most large network evolutions.
Back in the RAN market
While we wait, Dell’Oro did find that despite the broader state of the RAN market still being “underwhelming”, preliminary findings reveal that market conditions improved in the quarter, especially outside of China. “Regional coverage imbalances and monetisation challenges are nothing new. What sets this downturn apart is the uncertainty surrounding the capacity upgrade cycle that typically comprises a significant part of the RAN market in the new technology post-peak rollout phase,” said Dell’Oro RAN market research VP Stefan Pongratz.
Dell’Oro found that the revenue rankings did not change in 2024. The top 5 RAN suppliers based on worldwide revenues are Huawei, Ericsson, Nokia, ZTE, and Samsung. NTT Docomo’s Open RAN joint venture with NEC, Orex Sai, announced some network wins at MWC so it will be interesting if NEC can move the needle on market share. The top 5 RAN suppliers based on revenues outside of China are Ericsson, Nokia, Huawei, Samsung, and ZTE.
Interestingly, revenue shares changed in 2024 – Huawei and Ericsson improved their revenue shares while Nokia, Samsung, and ZTE’s RAN revenue shares fell. Dell’Oro said the fundamentals that shape the RAN market have not changed, and the long-term trajectory discussed in the most recent 5-year forecast still holds (0 percent CAGR over next five years).
The short-term outlook is mostly unchanged, with total RAN expected to remain stable in 2025 and RAN outside of China growing at a modest pace as conditions in North America and APAC excluding China improve.
