CSPs expect to play a broader role in AI connectivity, especially in providing high-bandwidth wavelength services for enterprise customers
In sum – what to know:
Long-haul networks hit hardest – Just over half (52%) anticipate AI will account for more than 30% of long-haul traffic in that timeframe, and nearly one-third (29%) believe AI will surpass the 50% mark.
Enterprise as a key driver – Three-quarters (74%) expect enterprises — rather than hyperscalers or cloud providers — to generate the largest share of new demand over the next three years.
Most CSPs aren’t ready – Network readiness is a work in progress. Only 16% of respondents say their optical networks are “very ready” for AI traffic today. Five percent say they are not ready at all.
Ciena and Heavy Reading (now part of Omdia) have released new global research highlighting how artificial intelligence (AI) is poised to dramatically increase traffic demands on metro and long-haul networks over the next three years — with communications service providers (CSPs) expecting AI to become a major driver of network growth.
The study, based on a custom survey of 77 CSPs spanning fixed, mobile, converged network operators, and cable providers, set out to understand how AI applications and workloads will impact capacity, service opportunities, and network readiness.
Metro vs. long-haul: Where AI traffic will hit hardest
The results reveal that CSPs expect AI’s share of total network traffic to grow significantly. In metro networks, 18% of respondents predict AI will account for more than half of all traffic by 2027, while nearly half (49%) expect it to exceed 30%.
For long-haul networks — which connect cities, data centers, and cloud regions — expectations are even higher. Just over half (52%) anticipate AI will account for more than 30% of long-haul traffic in that timeframe, and nearly one-third (29%) believe AI will surpass the 50% mark.
“The rapid rise of AI applications — from large-scale models to cloud AI services and edge-to-core workflows … are set to become major drivers of both local and long-haul network traffic,” said Sterling Perrin, Sr. Principal Analyst, Heavy Reading. “For metro networks, where AI will compete with video, web, and IoT traffic, the projected growth is striking. With AI expected to take an even larger share of long-haul capacity within three years, it’s clear that AI data flows, including those used for training and inference, will put unprecedented demands on CSP networks.”
Opportunities: High-bandwidth wavelength services lead the pack
CSPs expect to play a broader role in AI connectivity, especially in providing high-bandwidth wavelength services for enterprise customers. Half of respondents ranked these services — spanning 100G, 400G, and 800G rates — as the top growth area tied to AI over the next three years.
North American operators are particularly bullish, with nearly two-thirds projecting AI will make up more than 30% of both metro and long-haul traffic. By contrast, just over one-third of respondents in the rest of the world (RoW) predict similar AI-driven contributions.
Demand for dark fiber ranked lower in comparison, with only 25% of respondents expecting it to see the most growth from AI. Instead, many operators believe high-bandwidth wavelength services will overtake dark fiber as the dominant transport option for AI — especially as the market transitions from AI model training dominated by hyperscalers to enterprise adoption and day-to-day usage.
Enterprises seen as key AI traffic drivers
When asked which customer segment will drive the most AI-related traffic growth, CSPs pointed squarely to enterprise customers. Three-quarters (74%) expect enterprises — rather than hyperscalers or cloud providers — to generate the largest share of new demand over the next three years.
“Dark fiber builds for hyperscalers are an early win for some operators,” the report noted. However, it added that as AI adoption broadens, wavelength services are expected to become the dominant connectivity model for enterprises seeking AI capabilities.
Challenges: Capex and strategy top the list
While CSPs see the AI opportunity clearly, they also acknowledge that network readiness is a work in progress. Only 16% of respondents say their optical networks are “very ready” for AI traffic today, while 39% describe themselves as “ready” but still requiring upgrades. Another 40% say they are only “somewhat ready,” and 5% admit they are not ready at all.
The top three challenges cited in preparing for AI were capital expenditure constraints (38%), go-to-market and business strategy alignment (38%), and network management complexities (32%).
Geographic readiness gap
The survey found a notable geographic divide in network readiness perceptions. While 68% of RoW respondents say their optical networks are ready or very ready for AI, just 44% of North American operators feel the same. This gap likely reflects North America’s higher traffic expectations — and its greater concerns over capacity shortfalls — compared to other regions.
“Whether RoW CSPs are truly prepared for AI or simply are not yet seeing the traffic impacts will become apparent in the next one to two years,” the report observed.
A narrow window to prepare
The overarching message from the research is that AI-driven network demands are coming fast — and CSPs have a limited window to prepare. While many operators are optimistic, the report stresses that “it is very early days” and that the true traffic impacts of AI have yet to be felt.
The authors conclude that now is the time for CSPs to build strong business cases, secure investment, and accelerate optical network upgrades to position themselves for the coming AI traffic boom.
Table of Contents
