When we talk about AI, the conversation usually gravitates toward models, graphics processing units (GPUs), data center fabrics, breakthroughs, and productivity gains. But there’s a quieter question that will shape the next decade just as profoundly.
What happens to the wide area network?
Together with colleagues and partners, we’ve been studying something that hasn’t received enough attention yet: how AI—and especially agentic AI—is reshaping global wide area network (WAN) traffic patterns. Not in theory, not in hype cycles, but in measured data from service provider production networks, empirical testing of AI traffic characteristics, and forward modeling to establish a repeatable framework that can track traffic evolution over time.
AI network traffic is already reshaping infrastructure needs
What we are seeing is clear: AI isn’t just adding traffic. It’s changing the shape of traffic.
This is precisely why we wrote the AI Impact on Wide Area Networks report.
In the report, we identify some of the key differences in how AI traffic behaves compared to regular web transactions, particularly how inference-heavy communication paths suddenly become mission critical. Agents operate at machine speed instead of human speed, and that changes everything.
If AI models are the “brains” of this new era, then networks are the nervous system, and when autonomous agents begin to act, decide, and transact on behalf of humans at scale and machine speed, that nervous system of connectivity must be ready.
The goals of the AI Impact on Wide Area Networks report
Our intention with the report is not to predict distant sci-fi futures or summarize what everyone already suspects about AI. Instead, we want to begin a structured, data-driven conversation about:
- How AI inference traffic compares to non-AI web traffic at the network transport level
- What happens when agentic AI becomes embedded in enterprise workflows
- How consumer AI adoption changes internet growth curves
- Why traditional network planning assumptions may no longer hold
What sets the report apart is that it’s based on real-world traffic data, including an early lens on agentic AI traffic (currently small but growing fast) that lets us see and measure a new class of AI network traffic and understand the implications.
Unlike forecasts based on models alone, this report measures live AI inference traffic across real production networks—revealing how AI and agentic AI are reshaping infrastructure.
The report tracks the behavior of AI inference traffic flows over real production networks with controlled experiments to identify the network behavior and characteristics of AI applications, as well as modeling based on industry data.
The goal is to establish a repeatable measurement framework and baseline to track AI traffic evolution and forecasting on an annual basis to shed light and help network leaders make decisions.
AI adoption will have a compounding effect on traffic
AI adoption is accelerating at an unprecedented pace. Enterprises are embedding agents into core workflows, consumers are beginning to rely on autonomous AI assistants, and the compounding effect on traffic growth, symmetry, latency expectations, and critical path resiliency cannot be ignored.
The report estimates that by 2035, AI inference will represent 25% of all network traffic. This transformation will occur primarily between 2029 and 2032, when agentic AI adoption is projected to experience its most pronounced increase.1
AI inference traffic is expected to drive 63% additional growth compared to projection without the impact of AI because of the multiplying effect of AI applications. More insights and detailed analysis can be found in the report.
Critical infrastructure planning needed
AI will not just increase traffic volume—it will change traffic shape, symmetry, duration, and criticality. AI inference paths will become strategic network assets, requiring higher resilience, greater observability, and differentiated treatment, including quality of service and path security.
For service providers, network architects, and digital infrastructure leaders, the real risk is not that AI traffic will appear overnight. The real risk is assuming it behaves like everything else when it doesn’t.
The networking industry needs shared visibility, continuous measurement, and updated models to prepare for what’s coming over the next 10 years. This report marks the beginning of that effort.
If you are planning capacity, designing architectures, or defining strategy for the next decade, this conversation isn’t optional—it’s foundational. While AI inference is perceived as mostly a compute or GPU problem, the insights in the report indicate that as inference evolves, the networking part is becoming more relevant. For those who understand networks, that is not a challenge—it’s an opportunity.
AI is creating a new class of network traffic. We can now see, measure, and understand it. We invite you to read the report, challenge the assumptions, and join us as we continue this research journey. AI is already transforming software development and business processes, and quietly, but just as profoundly, it will transform the network as we know it today. See the highlights in this infographic or download the full report today.
Prepare your network for the AI-driven future
Explore detailed findings, methodology, and strategic recommendations for network operators as AI adoption accelerates through 2035. Download the AI Network Impact report.
- Estimates are based on extrapolation of the trajectory based on current observations. A faster pace of adoption and higher volumes cannot be ruled out. As more data becomes available, we will adjust future growth projections accordingly.
