AI Will Monetise 5G By Making It Cheaper

May 2026

Telecoms & Connectivity

Ardit Ballhysa

Senior Research Analyst

The arrival of 5G certainly brought with it noticeable advancements over 4G, such as higher speeds, lower latency, and improved capacity. Despite these advancements, mobile network operators (MNOs) have struggled to monetise 5G, as most consumers are unwilling to pay higher fees. Whilst enterprises have contributed more to monetising 5G than consumers - by paying the premium to capitalise on the technology’s benefits - the adoption of advanced 5G applications by enterprises has remained slower than initially anticipated by operators. Years after the commercial introduction of 5G, MNOs are still seeking ways to monetise the massive expenditure poured into the development of 5G and realise a meaningful return on their investment.

Artificial Intelligence-Radio Access Network (AI-RAN) is being positioned as a solution to this, as it repositions operators away from being pure connectivity infrastructure providers into becoming distributed AI and compute platforms. This will enable operators to support new enterprise workloads and automation capabilities, while also allowing them to offer edge computing services. However, one would be remiss not to notice the similarities between the promises of AI-RAN and those of 5G when it was under development.

AI-RAN and the Search for New Telecom Revenue Streams

Historically, RAN infrastructure relied mostly on static configuration and manual optimisation. Embedding AI into the RAN provides a far more dynamic operating model, in which networks can continuously - and autonomously - adapt to changing traffic conditions, application demands, and operational requirements in real time.

AI inference, the process of running trained AI models to generate real-time outputs at the network edge - such as identifying anomalies or automating industrial processes - is one of the most significant capabilities of AI-RAN. This is because rather than sending data back to centralised cloud datacentres, operators can process AI workloads closer to the end user through edge infrastructure embedded within the network. This creates the benefits of reducing latency and improving response times, as well as lowering bandwidth requirements. All these are extremely beneficial to enterprises; positioning AI inference as a leading functionality that will make 5G easier to monetise.

For instance, manufacturing facilities could deploy AI-powered quality control systems with near real-time responsiveness, while logistics companies could use AI-enabled tracking and predictive analytics across distributed operations. Operators will have the ability to charge higher fees in such instances; offering a new revenue stream for their 5G services.

Aside from providing new revenue streams, AI-RAN also has the potential to reduce operational expenditure (OPEX); owing to its autonomous capabilities. For example, AI-driven energy optimisation can dynamically place underutilised network equipment into low-power states during periods of reduced demand; potentially lowering energy consumption across 5G networks. AI-based traffic orchestration can allocate spectrum and network resources more efficiently based on application priority and real-time congestion patterns. Furthermore, predictive maintenance systems can potentially identify faults before outages occur; reducing downtime and operational disruption.

Costs Associated with AI-RAN

Deploying AI inference and training AI models requires intense computational power. This has led to NVIDIA becoming the leader in partnering with - and supplying its graphical processing units (GPUs) to - RAN vendors and operators, as can be witnessed through recent announcements from major vendors. However, GPUs, and in particular NVIDIA’s GPUs, are expensive; necessitating further spending from operators in the hope that they can extract a return from their initial 5G investment. But, herein lies the concern for operators, especially those that have racked up significant debt from investing in 5G. Asking operators to increase their spending, at a time when many are already facing pressure to monetise prior 5G investments, is not a clear-cut decision, especially as many operators remain wary of AI-RAN’s ability to generate significant revenue streams or reduce OPEX.

A more cost-effective alternative to NVIDIA’s GPUs is Intel’s Xeon 6 System-on-Chip (SoC), which offers the ability to run AI inference directly on the SoC without separate AI accelerators. The downside to deploying Intel’s solution, however, lies in its limited capacity to handle multi-modal neural networks or generative AI network agents at the edge to continuously monitor and self-heal network anomalies. This limitation will become even more pronounced with the onset of 6G, which will considerably increase network congestion. This means operators must decide whether to invest heavily in NVIDIA’s solutions to future-proof themselves, or opt for Intel’s more cost-effective solution; knowing that they may have to reinvest again within the next five years as 6G roll-out accelerates. 

Despite the complications around deployments and cost, operators must invest in AI-RAN to maintain or gain significance in their respective markets. Many of the leading operators globally are investing in incorporating services beyond traditional connectivity, and into cloud infrastructure management that is virtualised within an open ecosystem and driven by AI. This industry trend will continue and accelerate over the years; putting pressure on operators that have not begun following suit to do so in order to remain competitive.

The Near-term Benefits of AI-RAN

The number of AI applications requiring real-time inference close to the edge is increasing; providing opportunity for operators to monetise their investments in edge computing infrastructure. However, the most noteworthy near-term benefit to operators of investing in AI-RAN are the OPEX savings that will result. This is due to AI-based optimisation, which can dynamically reduce power consumption while maintaining - and in some instances improving - network performance. Furthermore, the autonomous capabilities of AI will play an important part in reducing energy costs associated with larger and more complex networks, which will only be intensified with the onset of 6G.

While AI has the potential to create such savings, the costs and energy requirements of running GPUs - in the cases where operators opt for them - must be considered, as they may outweigh any potential savings. Moreover, AI infrastructure itself is expensive to deploy and operate. To maximise returns from incorporating AI into their networks, operators must assess when and where to deploy AI; making sure to prioritise regions and environments where it will be utilised, as opposed to adopting it across their entire network under one big investment. It may also be more prudent for operators to initially invest in Intel’s Xeon SoC; owing to its affordability relative to NVIDIA’s GPUs, and upgrade to the latter only when there is a clear need and strong potential return on investment.