Nokia projects that AI will account for 44% of global network traffic by 2035, with the growth bringing more demands for uploads and rapid responses. The forecast points to a network shift: AI systems will need infrastructure built for machines talking back, not just people downloading more things.
Nokia Watch analysis
What happened
Nokia’s Global Network Traffic Report projects monthly global traffic will nearly quadruple from 2026 levels, reaching almost 3,789 exabytes by 2035. The report says autonomous AI agents and connected machines will increase demand for uplink capacity and real-time responses.
Nokia projects agentic AI traffic will grow to 160 times its 2026 level by 2035, while traffic from physical AI, including robots and vehicles, will reach 105 times that level. More than 90% of physical AI traffic is forecast to travel over the network uplink. Advanced Television’s report summarises the findings and Nokia’s projections.
Key findings
- AI’s projected share
Nokia forecasts AI will drive 44% of global network traffic in 2035. - A bigger job for uplinks
More than 90% of physical AI traffic is projected to travel from devices back into networks. - More processing near devices
Nokia estimates edge and metro infrastructure could serve 36% of AI-native traffic by 2035. - A fast-growing workload
Agentic AI traffic is forecast to reach 160 times its 2026 level by 2035.
Why it matters
The familiar internet pattern is largely a download story. Robots, vehicles and AI agents that send information back and need quick responses change the shape of the job: networks need enough upload capacity, low latency and computing closer to where data is generated.
Nokia’s figures are forecasts, not a tally of traffic already travelling through networks. But they put a useful number on the infrastructure question: AI planning is not only about chips and data centres. The connection between them, and the devices using them, has to keep up too.
Our read
The striking part is not simply that AI traffic may grow. It is that physical AI could be so upload-heavy. A network built mainly to deliver content to users may need a rethink when more of its customers are machines sending back what they see and do. Nokia has an obvious stake in networks being considered essential to the AI build-out, so these projections deserve to be read as a forecast, not a neutral measurement of the future.
What to watch
- Whether operators increase investment in uplink capacity and lower-latency networks.
- How much AI processing shifts towards edge and metro infrastructure.
- Whether measured traffic growth tracks Nokia’s forecasts as agentic and physical AI adoption develops.
Discussion spark: If AI systems send much more data back across networks, should operators prioritise uplink capacity and edge computing over faster downloads?
Sources and evidence
- Forecast: AI nearly half of global network traffic in 2035 | Advanced Television (4 November 2026, 00:00 UTC)
- Netflix rolls out AI Discover | Advanced Television (4 November 2026, 00:00 UTC)
not affiliated with or endorsed by Nokia