Artificial intelligence is shaping more than intelligence alone—it’s redefining how much bandwidth households and businesses will require in the near future. The Bandwidth Hawk notes that AI isn’t a temporary spike; it’s a sustained driver of demand that will reshape deployment plans and network architectures for years to come.
Consider modern videoconferencing: a typical work call might run at about 2 Mbps downstream and 0.5 Mbps upstream. When an AI-powered note taker is used, those figures can roughly triple, because the system streams audio to the cloud for transcription and delivers text back to the screen. That means per-user bandwidth could rise to around 5 Mbps downstream and 2–3 Mbps upstream, effectively multiplying demand in real time.
As video quality shifts from HD to 4K, bandwidth needs grow even more. With AI-driven processes layered on top of existing streams, the target of 20 Mbps per user by 2035 in the United States becomes more plausible, but only barely. And looking two decades out, the demand could reach about 80 Mbps per user, a level that would stress the BEAD target’s capacity to keep pace without substantial edge and fiber upgrades.
Beyond personal use, future homes could feature semi-autonomous robots for care, maintenance, and accessibility. Robotic tasks—from routine housework to complex repairs—will demand high processing power either in the cloud or at the edge, driving transient spikes that fiber-only backbones must absorb. The same trend extends to remote education, healthcare, and even autonomous vehicles sharing data at scale in city neighborhoods.
The Bandwidth Hawk emphasizes that the growth in traffic is not just about more devices, but about the type of activities AI enables. Global internet traffic already surged from 275 exabytes per month in 2020 to about 550 exabytes in 2025, with annual growth rates persisting in the high teens. The risk is not merely reaching a ceiling of bandwidth—it’s ensuring reliability, low latency, and resilient connectivity at the scale AI will demand. AI data centers, by necessity, require robust fiber trunks and edge networks, creating new deployment opportunities even as they challenge the sufficiency of today’s nominal BEAD targets. The message is clear: expect broader deployments, more fiber, and more intelligent routing to stay ahead of AI-driven demand.
