Nokia Sees AI Traffic Driving Fourfold Growth by 2035

By Daniel Okafor

Nokia forecasts global network traffic will nearly quadruple between 2026 and 2035, reaching 3,789 exabytes per month, with AI-native and AI-influenced applications generating 1,681 EB, or 44% of the total.

Nokia’s 2026 Global Network Traffic Report identifies generative AI, agentic AI, autonomous vehicles, industrial robots, connected cameras and fixed wireless access as the principal drivers through 2035. The report provides forecasts, not subscriber, coverage or measured throughput results.

Nokia projects AI-related traffic will grow at a 28% compound annual rate, compared with 12% for non-AI traffic. Nokia expects that mix to require more capacity across mobile access, fiber broadband, transport infrastructure and cloud connectivity.

What will AI add to global traffic?

• AI-native traffic will rise 60 times from its 2026 level to 987 EB per month in 2035. It will represent 26% of global traffic and grow at a 58% compound annual rate.
• Cloud AI workflows will account for 548 EB per month. Enterprise access AI will add 258 EB, and consumer access AI will add 181 EB.
• AI-influenced traffic, including extra consumption from recommendation systems, will contribute another 694 EB per month. Together, these categories explain the opening 1,681 EB estimate.

Nokia forecasts agentic AI will expand 160 times between 2026 and 2035, reaching 585 EB per month at a 76% compound annual rate. Cloud workflows will produce 462 EB and agentic access 123 EB.

Those systems connect agents, databases, software tools and external services through multiple interactions. One user request can therefore generate traffic across cloud and transport networks.

Nokia forecasts physical AI traffic will rise about 105 times to 67 EB per month by 2035, growing at 68% annually. The device forecast includes 17.1 million humanoid robots, 6.8 million commercial drones, 2.5 million robotaxis and about 1.7 billion cameras.

Continuous connectivity will support fleet coordination, remote operations, monitoring and cloud analytics. Nokia identifies private 5G, industrial connectivity, edge computing and enterprise services as the commercial outlets for this traffic.

Why does AI move traffic upstream?

Nokia expects AI applications to challenge the traditional “download-heavy design” of mobile networks. More than 90% of physical AI traffic will originate in the uplink by 2035.

Generative AI voice sessions can also place 90% of their traffic in the uplink. Smart glasses streaming video to cloud AI can produce an uplink-to-downlink ratio of 3:1.

In cells with high AI adoption, uplink could supply 24% of total access traffic. Operators will need stronger uplink capacity, better spectrum allocation and performance suited to real-time video, autonomous machines and industrial automation.

Where will AI traffic be processed?

Nokia expects the network edge and metro core to process 36% of AI-native traffic by 2035. The edge will handle 9%, while metro infrastructure will handle 27%.

Regional core and centralized cloud networks will retain the remaining 64%, especially for compute-intensive generative AI and enterprise services. Nokia calls the distributed model “the AI grid.”

Potential operator revenues include edge data centers, low-latency connectivity, AI inference infrastructure and enterprise cloud services. These are market opportunities in Nokia’s forecast, rather than booked revenue.

How much transport capacity will AI require?

Nokia estimates AI traffic exchanged with access networks will reach 1,134 EB per month by 2035. Counting traffic over individual network links will amplify that figure to about 3,720 EB, or 3.3 times.

More than seven in ten exabytes of that link traffic will cross communication service provider transport networks. Operators must therefore expand metro transport, backbone capacity and data-center interconnection alongside access networks.

What does growth mean for access and FWA?

Under Nokia’s moderate-growth scenario, global access traffic will reach 3,242 EB per month by 2035, a 15% compound annual rate. Conservative and aggressive scenarios produce 2,498 EB and 4,657 EB, respectively.

Consumer fixed access, including FWA, will remain the largest category at 2,121 EB. Consumer mobile traffic will reach 524 EB, while enterprise access traffic will reach 597 EB.

Video streaming, social networking, gaming, AI applications and enterprise digital transformation will continue to pull capacity through fiber, 5G and high-capacity transport networks.

Nokia projects consumer FWA traffic will reach 304 EB per month under the moderate case, more than one-third of total consumer cellular traffic. Combined consumer mobile and FWA traffic will reach 828 EB.

Asia-Pacific will generate more than half of global FWA traffic, supported by large deployments in markets such as India. North America will contribute about 17%.

Nokia identifies Reliance Jio and Bharti Airtel as operators that can use 5G FWA to raise broadband penetration and residential and enterprise revenue. Spectrum planning, densification and capacity management must keep pace with consumption.

Which enterprise markets will drive demand?

Nokia forecasts enterprise access traffic will reach 597 EB per month in the moderate scenario, growing at an 18% compound annual rate. The conservative and aggressive cases reach 382 EB and 908 EB.

Industrial IoT and extended reality will account for 55% of enterprise access traffic by 2035, exceeding traditional office traffic. Enterprise access AI alone will reach 258 EB per month, with a 59% compound annual growth rate.

Industrial IoT AI will supply 70% of enterprise access AI traffic, driven by cameras, robots, autonomous vehicles, sensors and industrial automation. That mix should increase demand for private wireless networks, reliable uplink, edge computing and managed enterprise services.

What should operators do next?

AI will shape telecom capital spending and network architecture through 2035. Operators must balance 5G capacity, fiber, FWA, transport and edge investments against workloads with different latency, reliability and uplink requirements.

Agentic and physical AI could create revenue beyond conventional connectivity, but the commercial test is converting traffic into sustainable returns. The risk is equally direct: capacity and edge spending may not translate into predictable revenue.

Before 2035, operators must show that uplink expansion, distributed computing and enterprise services can turn Nokia’s traffic model into paid demand.