As artificial intelligence (AI) is infused into more and more applications, operating systems and user devices, it starts to shape broadband network evolution. Network service providers have an opportunity to unlock new revenue by hosting and delivering AI applications cost-efficiently and reliably, as discussed in my previous blog post. But there are new challenges that may catch providers off guard if they’re not prepared.

How does AI impact network traffic?

This is the first question most network planners will ask and there are no easy answers. We asked Nokia Bell Labs researchers to investigate and present their findings in a Global network traffic report. Assuming moderate growth, they expect AI traffic to increase at a compound annual growth rate (CAGR) of 24% to reach 1,088 exabytes/month by 2030. This number accounts for consumer and enterprise traffic and adds up to about a third of global wide area network (WAN) traffic.

Figure 1. Global WAN AI traffic growth (exabytes/month)

A closer look reveals there are two types of AI traffic: direct and indirect. 

Direct AI traffic is generated by user or system interactions with AI applications and services. For example, consumers create direct AI traffic when they use generative AI, AI-assisted tasks, AI-powered gaming and extended reality (XR). Enterprises generate direct AI traffic with use cases that focus on improving operational efficiency and productivity, such as predictive maintenance, autonomous operations, video and image analytics, immersive multimedia applications and AI-enhanced customer interactions.

Indirect AI traffic results from AI algorithms that influence and increase user engagement by making it easier for users to find and compile relevant content. These algorithms spur traffic growth by providing personalized AI-driven recommendations for video content, social media, audio streaming and online marketplaces. 

So, the impact of AI is twofold: It drives higher demand for existing content (indirect AI traffic) and it enables new applications that are interactive, personalized and dynamic in nature and represent a high user value (direct AI traffic). Although direct AI traffic will not eclipse existing internet and video streaming in volume anytime soon, it is growing at a faster clip, demands more from the network in terms of service quality, reliability and security, and may add a significant source of revenues.

Why the broadband edge must evolve for AI

Currently, most IP networks have a broadband edge that is designed and dimensioned for best-effort internet and IP video streaming.  The broadband edge is an essential network demarcation point for managing subscriber access to digital content and services on the internet and in private clouds. It plays a crucial role in delivering, assuring and monetizing broadband services and must be flexible and adaptive to support multiple access technologies, higher data speeds, more users and more services. As we enter the AI era, the broadband edge must also be future-ready so it can take on new challenges and opportunities.

Video has been the dominant driver for faster access speeds and accounts for roughly 80% of internet traffic. Bigger screen sizes, higher-resolution content and on-demand viewing have all contributed to a steady increase in traffic. On average, broadband users spend nearly seven hours per day online, including more than two hours on social media, mostly watching video content. 

Although video and AI traffic is susceptible to fluctuations in bandwidth, packet loss and latency, most video content is created for public consumption, prerecorded and cacheable for on-demand viewing. When network congestion occurs, streaming protocols can dynamically lower the video resolution and use adaptive bit rate (ABR) streaming to prevent stalling. Live TV broadcast feeds can use IP multicast replication and buffering to enable efficient and reliable streaming to a large audience.

Figure 2. Broadband edge evolution for the AI era

None of the caching and congestion management mechanisms used for video delivery are viable for emerging real-time AI applications because most AI content is generated and consumed on the fly for personal use. Reliable, high-bandwidth transport is essential for making accurate AI inferencing decisions because it can maintain sufficiently high quality and resolution for image data and other sensory inputs. Low-latency data transport will ensure a responsive and seamless experience for real-time interactive applications such as robotics, live translation, auto-caption and XR, which must synchronize user inputs with AI-generated outputs. 

Delivering on deterministic bandwidth and latency guarantees will require more robust traffic engineering and congestion management capabilities, such as network slicing, hierarchical QoS and priority queuing. Integrating and distributing AI edge compute servers in the end-to-end data path will help to reduce data round-trip times, traffic congestion and security risks. 

It is imperative to secure the broadband edge against distributed denial-of-service (DDoS) attacks because they can severely impact bandwidth and latency-sensitive applications. Unfortunately, the use of AI has significantly enhanced the capabilities of malicious actors. As more insecure user devices connect at faster access speeds, there has been a steep increase in DDoS attacks. The Nokia Threat intelligence report 2024 noted a 166% year-over-year increase in DDoS attack volumes. IoT botnets were responsible for 60% of attacks. 

Journey to the edge of tomorrow

While traditional broadcast TV and IP video streaming is leveling off and may be reaching a saturation point, the use of new immersive media and intelligent applications powered by AI is projected to grow rapidly. Emerging AI applications generate less traffic than existing applications but will drive new demand for premium network services with better service quality, reliability and security. 

With a future-ready broadband edge, you will be able to navigate these changes and position your business for new growth opportunities in the AI era.

For more information, read our ebook or visit www.nokia.com/ip-networks/broadband-edge/.