Nokia forecasts that artificial intelligence will generate 44% of global network traffic in 2035, a shift it says will increase pressure on upload capacity and faster data processing.
Nokia’s estimate puts monthly traffic at almost 3,789 exabytes (EB) in 2035, close to four times the 2026 amount. AI would account for 44% of that total, according to the report. These figures are projections.
Traffic from autonomous AI agents and connected devices is expected to involve more data moving back toward networks and more exchanges that require prompt replies. Nokia says this differs from infrastructure designed chiefly to send data out to users.
Agentic AI refers here to autonomous software agents, while physical AI includes machines such as robots and vehicles that interact with their surroundings. Nokia identifies both categories as fast-growing traffic sources, alongside the consumer and enterprise activity included in its broader forecast.
For robots and vehicles using physical AI, Nokia expects uplinks — routes carrying data from devices into networks — to handle more than 90% of traffic in 2035. It projects agentic AI traffic to be 160 times its 2026 volume and physical AI traffic 105 times its 2026 volume, with both multiples starting from a small base.
The forecast also points to a change in where AI data is processed. Central cloud facilities are expected to remain responsible for most AI traffic, while edge and metro core systems nearer users and devices could handle 36% of AI-native traffic. Nokia ties that distribution to faster-response requirements, data sovereignty and more efficient models, and calls the network concept an AI grid.
On cellular networks, fixed wireless access (FWA) is forecast to account for more than a third of consumer traffic by 2035 as connections and usage expand. The projection places home broadband use alongside mobile services in Nokia’s expected traffic mix.
Nokia says it built the 2026–2035 forecast by modeling demand from consumers, businesses and AI. It compared that model with its lab work and expert assessments, while also drawing on outside research, app-use patterns, connected-device activity and data on agent workflows.
“AI is changing what we need from networks,” said Pallavi Mahajan, Nokia’s chief technology and AI officer. She said planning should account for machine demands as well as the amount of traffic.
