I haven’t talked about the growth of broadband usage for a while. However, I was explaining the exponential growth of broadband usage to somebody recently, and I suddenly realized an easy way for putting broadband growth into context.

The amount of data used by the average broadband user has been doubling roughly every three years since the advent of the Internet. This exponential growth has been chugging along since the earliest dial-up days, and we’re still seeing it today. Consider the following numbers from OpenVault showing the average monthly U.S. household broadband usage:

• 1^st Quarter 2018 215 Gigabytes
• 1^st Quarter 2019 274 Gigabytes
• 1^st Quarter 2020 403 Gigabytes
• 1^st Quarter 2021 462 Gigabytes

Average household usage more than doubled in the three years from 2018 to 2021. The growth happened at a compounded growth rate of 29% annually. That’s a little faster than the more recent past, probably due to the pandemic, but the compounded annual growth rate was around 26% in the decade before the pandemic.

What does this kind of growth mean? One way to think about broadband growth is to contemplate what growth might mean in your own neighborhood. Supposes you are served by DSL or by a cable company using HFC technology. If your ISP has the same number of customers in your neighborhood now as in 2018, the local network is now twice as busy, carrying twice as much traffic as just three years earlier. If your ISP hasn’t made any upgrades in that time, the chances are that you can already see some signs of a stressed network. Perhaps you notice a slowdown during the evening prime time house when most of the neighborhood is using broadband. You’ve probably run into times when it was a challenge making or maintaining a Zoom call.

This kind of broadband traffic growth might not seem like a big deal to the average person because they probably assume that ISPs are doing magic in data centers to keep things working. But any network engineer will tell that a doubling of traffic is a big deal. That kind of growth exposes the bottlenecks in a network where things get bogged down at the busiest times.

The most interesting way to put broadband growth into perspective is to look into the future. Let’s say that the historical 26% growth rate continues into the future. There is no reason to think it won’t because we are finding more ways every year to use broadband. If broadband keeps growing at the historical rate, then in ten years your neighborhood network will be carrying ten times more traffic than today. In twenty years it will be carrying one hundred times more traffic than today.

When you think of growth in this manner, it’s a whole lot easier to understand why we shouldn’t be funding any technologies with grant money today that won’t be able to keep up with the normal expected growth in broadband traffic. Looking at growth from this perspective explains why AT&T made the decision last year to stop selling DSL. Understanding the normal growth rate makes it clear that it was idiotic to give CAF II funding to Viasat. Expected growth might be the best reason to not give RDOF subsidies to Starlink.

I have nothing against Starlink. If I still lived in a rural area, I would have been one of the first people on the list for the beta test. But there are already engineers I respect who believe that the Starlink network will struggle if the company sells to too many customers. If that’s even just a little bit true today, then how will Starlink perform in ten short years when the traffic will be ten times higher? And forget twenty years—Starlink is at heart a wireless network, and there are no tweaks to a wireless network that will ever handle a hundred-fold increase in traffic. If Starlink is still viable in twenty years, it will be because it took the same path as Viasat and imposed severe data caps to restrict usage or else raised rates to restrict the number of customers on the network.

I take flak every time I say this, but if I was in charge of grant programs, I wouldn’t fund anything other than fiber. I can’t think of any reason why we would fund any technology that doesn’t have a reasonable chance to still be viable in ten short years when broadband usage will likely be ten times higher than today. I would hope that a government-funded network will still be viable in twenty years when traffic volumes are likely to be one hundred times greater than today. If we don’t get this right, then we’re going to be talking about ways to build rural fiber a decade from now when other technologies crash and burn.