Ensuring Telco Networks Are Ready for AI Monetization
Artificial intelIigence (AI) monetization opportunities are requiring telecom networks to become more innovative. Beyond new revenue from sovereign and enterprise AI, service providers use the technology to cut costs and improve reliability.
To support these goals, operators are modernizing transport equipment. These upgrades enable automated, adaptive network management – and stronger security.
James Blackman, global editor of RCR Wireless News, met with industry experts to discuss telco AI market developments. The group, which included speakers from ABI Research, Tupl, Red Hat, Keysight, and Verizon Business, gathered for the Supporting AI, Using AI webinar.
During the session, industry market strategy expert Stephen Douglas provided insight on AI traffic. Below, discover his advice on optimizing networks for AI traffic and monetization.
The telco AI monetization opportunity
Early adopters are eyeing monetization strategies to become AI enablers. They’re leveraging assets to offer more powerful core and edge localized services, including inferencing at the edge, cloud partnerships, and private secure slices.
Sound strategies use competitive assets as a foundation to monetize services. These include:
- AI networking as a service, providing optimized connectivity and scaling
- AI infrastructure as a service (for example, Graphics Processing Unit (GPU) as a Service, Inference at the Edge, and public cloud partitions)
- AI models as a service, managing foundational, custom, and industry-specific models
- Off-the-shelf and custom AI applications
Figure 1. Telco AI monetization opportunities
While networking is essential for AI monetization, it places complex new demands on existing network assets.
Transforming from a pipe to a performance platform
Early engagements with operators show that supporting AI traffic isn’t business as usual.
AI services – such as image generation, chatbots, and the Internet of Things (IoT) – have diverse transaction volumes, size, latency, and inference concurrency requirements.
This results in variable AI networking demands with distinct characteristics. Traffic can include unpredictable throughput intensities and bursty inference spikes across the network. Operators must manage high levels of concurrency at inference sites, and certain AI tasks require strictly controlled latencies.
These AI traffic patterns fundamentally shift the network’s role from a pipe transporting uniform traffic to a performance platform that must address variable and stringent requirements. The platform must be dynamically scalable, resilient, and highly performant.
Optimizing networks for AI
Operators are building and optimizing networks to support AI. They are investing and testing to ensure infrastructure and transport network capabilities can evolve to support novel AI traffic patterns and loads.
One optimization priority is transport network design for performance and scalability. Operators are upgrading wide area networks to 400 Gbps capacities. They’re also implementing network slicing for AI application traffic with segment routing, IPv6, and flexible Ethernet.
We also see optimized traffic engineering and routing to AI data center facilities at the edge, core, or in partner cloud networks, as well as optimized security designs.
Supporting AI traffic is complex. High throughput, low latency, determinism, minimal jitter, and low to zero packet loss are essential for AI performance. Networks must provide automated path failovers, elastic scaling to handle traffic bursts, quality of service (QoS), slice-specific service-level agreements (SLAs), and synchronous and parallel processes for concurrency inferencing.
The impact of AI on the transport infrastructure is broad, ranging from AI users to AI inferencing at the access/edge, tuning at the metro/regional level, and training at core/central data centers. These capabilities must also align vertically across the full-stack Layers one through seven.
Only then can the network become a monetizable product that offers services to enterprises and sovereign bodies.
How test and assurance ensure networks can support AI
To offer AI networks as a premium service to enterprises or to sovereign entities, network performance and precision must be guaranteed. Reliability is the currency of AI networking, enabling operator networks to become platforms for growth in the new AI economy.
Companies like Keysight help service providers execute robust test and assurance strategies. This helps ensure that networks built for AI and networks enhanced with AI are trustworthy and ready to support heavy investments.
A successful test and assurance strategy ensures performance under stringent real-time conditions, as well as:
- Validates AI efficacy, to build trust and transparency
- Ensures regulatory compliance and performance under stringent real-time conditions
- Protects sensitive data and infrastructure with continuous security and resilience testing
- Assures QoS by validating SLA compliance and providing proactive assurance that minimize business and legal risks
- Enables continuous improvement with actionable insights and feedback loops
Innovative methodologies include realistic emulation, use of synthetic test data, continuing test and automation, active testing, and non-functional testing.
Network digital twins provide realistic emulation and simulation to assess AI solutions and network behavior, performance, security, and efficacy ahead of live deployment. Synthetic test data simulates real-world traffic and data for controlled, repeatable network and system testing.
Continuous testing runs automated tests at every stage of the lifecycle, while active testing adds synthetic test traffic into the network to simulate network functions and usage patterns. This enables proactive monitoring of performance, SLAs, and AI-orchestrated network changes.
Non-functional testing evaluates how networks perform under real-world conditions by assessing performance, scalability, security, and reliability.
AI offers service providers monetization opportunities, as well as operational automation and efficiency. However, AI traffic is far more complex than traditional transport networks are optimized to handle. Operators are incorporating test and assurance plans in their AI networking strategies to ensure they’re ready for AI’s novel traffic and monetization potential.
For more insight, visit our blog on advancing AI opportunities at the telecom edge.