Nokia’s AI traffic forecast puts upload capacity on the planning agenda

Nokia expects AI to account for 44% of global network traffic by 2035. Its Global Network Traffic Report, released on 4 November, forecasts that total traffic will almost quadruple over the period. For managers planning AI projects, the useful question is what those workloads will need from a connection. ¹

The company anticipates growing demand for uploads and rapid responses as software agents and connected machines become more widely used. It projects that robots, vehicles and other physical AI applications will send more than 90% of their traffic upstream by 2035. Upload capacity is the ability to send data out; latency is the delay before a response arrives. ¹

That distinction gives a project team something concrete to test. Consider a proposed warehouse inspection system that sends camera footage for remote analysis. A connection that handles office downloads comfortably could still struggle when several cameras upload together. If an operational decision depends on the result, the time taken to return it also matters. This is an illustrative planning example.

The commercial implication is to include connectivity in the business case. Ask what information leaves the site, how much is sent at busy periods, and how quickly the system needs an answer. Establish what happens when the connection slows or fails. Processing some information locally may be one option to assess alongside additional capacity.

Nokia sells network infrastructure, and these are its long-term projections. Treat them as a reason to examine the assumptions in a specific project. Before approving a rollout, ask the supplier to demonstrate performance using representative data and the connections available at the intended sites. That gives the investment discussion a firmer basis than a headline download speed.