PARTNER CONTENT: Huawei highlighted new AI technologies will enable the telecoms industry to move towards higher-level network autonomy which is vital for the success of new 5G-Advanced (5G-A) networks.

Speaking at MWC Shanghai 24 on the topic of The Rise of 5G-A, Yang Chaobin, Board Member and President of ICT Products and Solutions at Huawei, said ICT convergence is the key driver behind higher-level network autonomy and outlined how its automated driving network (ADN) can boost 5G-A network productivity.

Generative AI has made significant breakthroughs across the world, positioning the technology as a foundational driver for network automation. “We are continuously integrating rapidly evolving AI technologies to improve the level of network automation,” he noted.

Meanwhile, on the network side, 3GPP release 18, the first standards unveiled for 5G-A, was officially frozen in mid-June, preparing the way for the widespread launch of networks.

In late March, China Mobile switched on the world’s first commercial 5G-A network, turning on the service in Shanghai, where nearly 150,000 households are already connected to the new technology. The operator plans to deliver 5G-A connectivity to more than 300 cities by end-2024.

Combined, the advances enable improved network O&M efficiency by transitioning from manual configuration to tool-assisted task automation, he said.

Data boost
From a services development perspective, AI is transforming information production, processing, transfer and exchange, bringing new opportunities in the mobile AI era.

At a time of declining data growth, the speed at which content such as AI generated video is growing is expected to spur a new round of network traffic growth.

China Unicom recently noted uptake of broadband services fell to just 2.3 per cent last year after growth peaked at about 30 per cent in 2014.

That is about to change. Huawei predicts mobile traffic growth will see a turning point in 2025, with a 20-times increase by 2030.

The number of AI assistant users is forecast to reach 1 billion by 2030, while AI is expected to be applied to tens of thousands of industry segments.

Yang argued these developments will place new demands on operators as they work to differentiate their offerings to enterprises, including end-to-end ultra-low latency and deterministic assurance. More advanced network automation capabilities will also be required. These cover user self-service capabilities, human-machine natural semantic interaction capabilities and self-closure capabilities for network service scenarios.

Capex vs opex
From the network investment perspective, over the past two decades, operators’ capex and opex have diverged.

As a percentage of revenue, capex is falling while opex continues to maintain a high share. Capital spending is predicted to drop to just 14 per cent of revenue in 2025, down from 50 per cent in 2020. Meanwhile, opex is on track to rise 68 per cent from 45 per cent over the same period.

By analysing the specific components of opex, Yang noted it identified large expenses, including installation and maintenance, electricity and network optimisation. These can account for up to 58 per cent of total expenses.

“We believe that introducing telecoms foundation models to build a new O&M mode is the key to reducing opex, which can overcome challenges like network complexity, precision and efficiency,” he explained.

At MWC Barcelona in February, the company launched the Huawei Telecom Foundation Model, which improves the capabilities of its ADN solution.

Flexible options
Based on the core capabilities of foundation models, Huawei can provide various applications, copilots and agents to reshape the O&M paradigm.

Its full-stack architecture models introduced to networks cover three layers:

  1. The business operation layer offers an intelligent window for operators to provide services for users, who can subscribe to and adjust services as required.
  2. The network O&M layer completes closed-loop O&M by acting as the intelligent brain of operation and the centre for foundation model training. It generates optimal O&M policies for the entire network.
  3. The network resource layer completes closed-loop resource management and executes policies.

Yang noted that efficient closed-loop network maintenance is implemented through intelligent analysis and inference.

Service rollout and monetisation are accelerated through intent interaction and automatic API generation. Focusing on high-value scenarios in network maintenance, experience assurance and service enablement, Huawei built five role-oriented copilots and five scenario-oriented agents.

The copilots provide natural semantic interaction, intelligent knowledge Q&A and O&M assistance capabilities, which help engineers reduce manual information query and collaboration between operators, lower the operation threshold and improve the operation efficiency of each of the five roles.

For five complex O&M scenarios, the agents can understand and break down complex tasks, invoke appropriate tools and APIs, and quickly and independently close tasks, which enables service quality improvement and revenue generation, O&M cost reduction and efficiency improvement.

Yang closed his speech by acknowledging the road to fully automated O&M processes is complex and requires collaboration across the entire industry, but underscored the significant successes accelerated by new AI capabilities.