I made a phone call last week, an actual voice call to an actual person, and noticed that I had not made one in perhaps three weeks. Most of my communication, including with the same person I called, happens via text — sometimes SMS, more often WhatsApp or iMessage. The phone call felt slightly anachronistic, as if I had taken out a fax machine.
This is not a new observation; everyone has noticed by now that voice calling has been in long-term decline. But there is a more specific version of the observation that gets less attention: the mobile network was built to carry phone calls, and phone calls are now a vestigial use case for the network. The infrastructure built for one purpose has been quietly repurposed for another, and the original purpose has nearly disappeared from the load.
The original mobile network
The first generation of mobile networks, deployed in the 1980s, was an analog system designed exclusively for voice. The 800 MHz spectrum allocated to mobile in the United States was specifically reserved for cellular telephone service. The technology was conceived as an extension of the wireline telephone system into a wireless context.
Even as the technology became digital with 2G, the central use case remained voice. SMS was a side feature added almost as an afterthought, using control channels that were originally intended for network signaling. The early data services — circuit-switched data, then GPRS — were slow, awkward, and clearly secondary to the voice traffic the networks were really designed to handle.
3G was the first generation that took data seriously, but voice was still the dominant use case for most subscribers through the 2000s. The famous battle between Apple and AT&T in the early iPhone era was largely about whether AT&T's network could handle the data load that smartphones generated. The network had been engineered to carry phone calls; smartphone data overwhelmed it.
The transition
By the time 4G LTE rolled out in 2010-2012, the architecture of the network had been redesigned around data. Voice itself became a kind of data — Voice over LTE, or VoLTE, treats voice calls as just another IP packet stream. The dedicated circuit-switched voice infrastructure that had been the core of the network became progressively less important and eventually was retired.
The transition was not visible to subscribers. From the outside, voice calling continued to work the way it always had. From the inside, voice was being demoted from "the thing the network exists to do" to "a particular application running over an IP network that is mostly carrying other things."
What the network actually carries
Modern mobile networks are dominated by video. Streaming video — Netflix, YouTube, TikTok, Instagram Reels — accounts for a majority of mobile data traffic. Web browsing, app data, software updates, and other miscellaneous traffic accounts for most of the rest. Voice calling is a small percentage of network load. SMS is even smaller.
The network engineers I have known sometimes refer to voice calling as "noise" in their capacity planning. It is not zero, but it is small enough that it does not meaningfully constrain network design. The capacity of the network is shaped almost entirely by video and high-bandwidth data applications.
This is a quiet inversion of the network's original purpose. The thing the network was built to do is now an afterthought. The thing that was an afterthought is now the load.
What this teaches about infrastructure
I find this an interesting case study in how infrastructure systems evolve. The mobile network was not redesigned for video; the mobile network was redesigned for higher-bandwidth data, and video happened to be the most demanding application. Each generation of network solved the immediate problem and inadvertently enabled the next generation of applications.
The same pattern shows up in other infrastructures. The internet was originally designed for academic email and file transfer; it now carries video calls and four-dimensional virtual environments. The interstate highway system was designed for defense logistics and intercity passenger travel; it now carries the just-in-time logistics of every product in every store.
Infrastructure built for one purpose tends to be repurposed by emergent applications its designers did not anticipate. The mobile network is a particularly clear case because the original purpose was so narrow and the actual current usage is so different.
What it suggests for the future
The natural question is what 6G will be repurposed for. The current consensus is that 6G will be designed around even higher bandwidths, potentially with integrated artificial intelligence at the radio layer, and with much tighter integration with satellite systems for ubiquitous coverage. The applications that drive 6G demand are presumed to be extended reality, autonomous systems, and machine-to-machine communication at scale.
It is possible all of this is correct. It is also possible that 6G will be designed for these applications and end up being repurposed for something else entirely — some new kind of consumer behavior or commercial application that does not yet exist. The history of the mobile network suggests that the actual use cases will surprise the designers.
A closing thought
I think about this when I make a rare phone call. The infrastructure that connects the call was built for that purpose, decades ago, by people who could not have imagined the things it would later be asked to do. The original purpose persists, vestigially, like the appendix in mammalian biology — present, occasionally noticed, no longer central.
The phone call goes through. The network barely notices. The thing it was built to do has become the smallest thing it does.
