Artificial intelligence is not only revolutionizing the software industry and data centers but also emerging as one of the primary drivers of network traffic growth in the coming decade. This is highlighted in the report “Optical Transport Networks for AI” produced by Ciena and Heavy Reading (Omdia), which warns that within just three years, nearly one-third of communication service providers (CSPs) believe AI will account for more than half of their long-distance traffic.
The study, based on a global survey of 77 fixed, mobile, converged, and cable operators, quantifies a trend that network engineers are already observing: the rise of AI applications—ranging from large-scale models to cloud services and edge-to-core flows—will demand unprecedented transport capabilities across both metropolitan and backbone networks.
AI: From the Lab to Network Traffic Domination
In metropolitan networks, 18% of CSPs expect AI to surpass 50% of total traffic by 2028, with nearly half (49%) anticipating it will exceed 30%. The projection for long-distance networks is even more ambitious: 52% expect AI to capture more than 30% of traffic, and 29% foresee it surpassing half.
Heavy Reading senior analyst Sterling Perrin summarizes:
“AI will directly compete with video, web, and IoT on metropolitan networks, but it will be even more dominant on backbone networks. Data flows for training and inference will require bandwidth and management capabilities never seen before.”
The Business Behind AI: High-Capacity Optical Services
The report indicates that operators foresee a more active role in connectivity for AI, especially in the enterprise segment. 50% identify high-capacity wavelength services (100G, 400G, up to 800G) as the main growth opportunity related to AI over the next three years.
In comparison, only 25% expect growth from dark fiber. Additionally, 74% believe enterprise clients will drive the largest traffic increase, ahead of hyperscalers and public cloud providers.
Insufficient Preparation for the Data Tsunami
While the business opportunity is clear, technical readiness lags behind. Only 16% of operators consider their optical networks “very prepared” to meet AI demands. Another 39% feel “ready” but with work to do, 40% admit their preparation is partial, and 5% say they are not ready at all.
Key obstacles include investment restrictions in CAPEX (38%), commercial and market strategies (38%), and network management (32%).
A Race Against the Clock
The message from the report is clear: the window to adapt optical infrastructure for the new AI era is closing rapidly. Operators who don’t upgrade their optical transport capacity and network management risk losing competitiveness against rivals capable of offering low latency, optimized routes, and enough capacity for massive data flows.
In a context where AI generates traffic not only in data centers but also in edge-to-core and multi-cloud flows, transport network design will become a strategic necessity, not just a technical concern.
Frequently Asked Questions (FAQ)
What is a high-capacity wavelength service?
It’s a dedicated optical connection that transports data at speeds like 100G, 400G, or 800G, used for critical applications requiring high bandwidth and low latency.Why does AI increase network traffic so much?
Because training and inference processes for models require moving huge volumes of data between data centers, clouds, and edge devices.What role do operators play in this ecosystem?
Beyond providing connectivity, they can offer AI-optimized services like dedicated routes, private optical networks, and peering agreements tailored for AI workloads.What is the biggest challenge in adapting to AI demand?
Balancing infrastructure investment, optimizing network management, and developing competitive commercial offerings before demand surpasses current capacity.
via: ciena