Nokia Forecasts AI Will Drive 44% of Network Traffic by 2035

Nokia expects AI workloads to account for 44% of global network traffic by 2035, with total traffic nearly quadrupling from 2026 levels to almost 3,789 exabytes per month. The projection comes from the vendor's Global Network Traffic Report (2026–2035), published on 4 November 2026, and it points to a structural problem for operators: networks engineered primarily for downloads will have to serve machines that upload heavily and demand real-time responses.

What does the forecast change for network design?

The short answer is the uplink. Physical AI applications, defined in the report as robots and vehicles that interact with their surroundings, are projected to generate more than 90% of their traffic over the network uplink by 2035. Agentic AI traffic, starting from a small base, is expected to reach 160 times its 2026 level in 2035. Physical AI traffic is forecast to reach 105 times its 2026 level over the same period.

These workloads differ in character. Some generate bulk uploads; others require continuous, low-latency connections. A network built around streaming and browsing cannot serve both without redesign.

Where the processing goes

The report also projects a shift in where AI data is processed. Most AI traffic is expected to remain within the central cloud. However, edge and metro core infrastructure is projected to serve 36% of all AI-native traffic by 2035 as agentic and physical AI applications gain prominence.

Three forces drive that distribution, according to Nokia:

• more efficient AI models
• low-latency requirements
• data sovereignty needs

Together they pull intelligence and inferencing closer to users and devices. Nokia labels the resulting distributed infrastructure the AI grid, a term that reflects the vendor's own product positioning as much as it describes an operator requirement.

Fixed wireless access keeps growing

Beyond AI, the report projects that fixed wireless access (FWA) will generate more than one-third of consumer cellular traffic by 2035 as connections and usage grow. That expands the role of cellular networks in home broadband alongside traditional mobile services, a trend already visible in FWA-heavy markets such as the United States and India.

Methodology

Nokia's model is bottom-up. It combines application usage, connected-device activity and agentic workflows with third-party research, expert review and Nokia lab testing to assess traffic growth and characteristics, network distribution and potential uplink constraints. It is a vendor forecast, and Nokia sells the AI-ready infrastructure the forecast argues for; the traffic figures nonetheless rest on a documented modelling approach rather than marketing assertion alone.

The operator message

Pallavi Mahajan, Chief Technology and AI Officer at Nokia, framed the stakes directly. "AI is changing what we need from networks, not just how much traffic they carry," she said. "Networks built for traditional internet applications will increasingly have to support machines that place new demands on the uplink and require real-time responses."

Mahajan added that the next phase of AI will depend as much on the network as on compute. "Network readiness must therefore be a central part of AI planning, not an afterthought," she said, noting that Nokia is building the distributed, AI-ready networks it says the shift requires.

For operators, the practical implication spans spectrum policy, uplink capacity investment and edge build-out over the coming decade. Nokia's timeline runs to 2035, with the steepest AI-driven growth expected in the second half of the period.