AI WAN: Leading IP Networks into the Intelligent Era

As the telecoms industry hurtles towards the digital world, demand is already growing for emerging bandwidth-intensive applications and services such as artificial intelligence (AI).

Telecoms operators worldwide have already set their sights on 5G-A to handle that demand, but their IP networks must also evolve to handle the exponential growth in traffic demand and deliver the essential quality of experience (QoE) customers expect.

Many operators have already identified AI as a crucial element of IP network evolution, and are developing strategies to converge AI into their networks. The World Broadband Association is currently developing the Net5.5G standard that aims to define next-generation IP for intelligent bearer networks. The next stage of Net5.5G evolution is AI WAN.

Huawei Technologies unveiled its AI WAN solution at Mobile World Congress 2025 in Barcelona. The solution enables carriers to build networks with optimal TCO, expand service boundaries, improve operations efficiency, enable scalable new services and deliver QoE levels that will give them an edge over the competition.

According to Leon Wang, President of Huawei's Data Communication Product Line, “AI WAN comprehensively empowers IP networks using AI in the Net5.5G era.”

Challenges in IP network evolution

To understand why AI is essential to IP network evolution, it’s worth looking at some of the key challenges operators face that legacy networks struggle to overcome.

For a start, many operators faced with flat or stagnant revenue growth have tried to leverage QoE to charge more for services, but this has proven difficult to execute. For example, a European operator launched a video acceleration service in 2022, where customers could pay extra for "zero buffering" of videos such as Netflix. However, the extra load of network traffic encryption ended up increasing the buffering rate from 2% to 60%. Within half a year, the user churn rate was 19%.

Another challenge is that new services can’t be scaled up quickly. For example, a carrier in the Middle East launched cloud-based gaming and home-monitoring services in 2022, but since then, less than 20% of its home broadband customer base have adopted the service, while 40% of users said they experienced frame-freezing due to network delay.

A more fundamental challenge for operators is rising opex. Globally, 68% of revenues on average will have to cover opex costs in 2025, and for some operators it will exceed 80%.

To address these problems, operators such as Orange, Deutsche Telekom and Vodafone Group have released AI strategies to explore integrating AI into their networks, and this trend is expected to accelerate – especially as the training costs and usage costs of large language model (LLM) technologies has come down significantly by a factor of 10.

AI WAN directs Net5.5G evolution

The question then becomes how to go about integrating AI into networks. This is where Net5.5G comes in. Net5.5G serves as the foundation for digital IP bearer networks and defines the sustainable evolution of IP network infrastructure in the era of 5G-A and ubiquitous AI computing.
The WBBA is collaborating with operators to promote the integration of IP networks and AI around the Net5.5G standard. By leveraging AI in network operations, experience assurance, and network equipment, the WBBA aims to achieve comprehensive enhancements in network capabilities, value, and efficiency.

A key component of this is AI WAN, which serves as the IP bearer network for Net5.5G. Huawei’s AI WAN Solution features a three-layer architecture:

• “AI routers”, which integrate millisecond-level flow reporting, high-accuracy flow identification, and efficient security protection engines, building an AI WAN foundation for intelligent capabilities.
• “AI new connections”, which enable flow-level scheduling to meet the diverse network requirements of various applications, allowing carriers to offer a wider range of value-added services.
• “AI new brain”, which improves O&M efficiency by leveraging Huawei’s Network Digital Map (digital twin) and Network Foundation Model to create network AI agents that assist carriers with network functions such as online change simulation, fault diagnosis, and fault handling.

The AI routers are the foundation of the AI WAN solution. According to Huawei, an effective AI router should possess three key engine capabilities:

• Traffic collection engine: because data serves as the foundation for both training and inference, the AI router should support precise traffic collection and preprocessing at the millisecond level for ultrafine granularity, enabling it to detect things such as attack flows and anomalous micro-bursts.
• Traffic identification engine: An AI router needs to be able to identify business traffic – including diverse flow characteristics such as encrypted flows, mutated application flows, and elephant flows – with an accuracy rate of no less than 95%. This enables business data flows to be scheduled flexibly to provide new services.
• Security protection engine: This engine enables the AI router to defend against APT attacks, prevent implantation of malware, avoid the theft of confidential information and account impersonation, and detect abnormal behaviours within routers, with a 95% detection rate for known threats.

New growth and new market spaces

Huawei’s Leon Wang notes that none of this is theoretical – Huawei has already road-tested its AI WAN solution to demonstrate the technology’s benefits, and has made some innovations with some operator customers that bring new value to operator business scenarios such as ToM, ToH, and ToE (to-mobile, to-home and to-Enterprise).

For example, AI WAN’s traffic collection engine enables it to accelerate ROI for individual services by utilising predictive operations to efficiently identify sites with congested traffic, which enables carriers to make targeted investments. MTN South Africa’s base stations – leveraged AI WAN's millisecond-level traffic collection and minute-level prediction capabilities to identify the 10% of base stations impacted by severe link congestion, which was compromising traffic experiences of users. Result: dataflow of usage (DOU) rose by 25%, while traffic grew by 15.4%.

AI WAN’s ability to accurately identify encrypted flows via its traffic identification engine can unlock new opportunities for carriers to monetise differentiated experiences. Macau SAR carrier CTM recently partnered with Huawei to optimize network services using an AI computing engine that resulted in a dramatic reduction in game latency and thus a significant improvement in the customer experience. Another operator worked with Huawei to conduct experience-centric operations using AI to analyse and spot poor QoE, enabling it to provide deterministic cloud-network services. The solution not only significantly enhanced user experience but also attracted a substantial number of new cloud broadband package subscribers.
Meanwhile, AI WAN’s security protection engine enables new security service offerings for enterprises such as value-added intelligent flash defence services, which can accurately identify attack flows. Wang said this could increase enterprise business revenue by 35%.

Also, AI WAN provides elastic and lossless transmission capabilities, helping carriers expand new integrated computing-network services for businesses. Huawei and China Telecom Shanghai have done exactly this, using the AI WAN Solution to create an end-to-end 400GE intelligent computing network for computing WAN scenarios. Key technologies, such as lossless WAN, were employed to enable efficient transmission of computing power services, enabling China Telecom to offer efficient computing power leasing and computing network services to industry customers.
This is just a sample of what AI WAN can do to help operators effectively integrate AI into their IP networks and lead them into the era of truly intelligent networks, said Wang.

“Huawei will continue to focus on network intelligence innovation, developing industry-leading products and solutions,” he said. “We will work closely with partners to build AI WAN, accelerate carriers' service growth, and jointly move toward the Net5.5G intelligent network era.”