One of the more interesting rules-of-thumb in the industry is Nielsen’s Law of Internet bandwidth, which states that:

A high-end user’s connection speed grows by 50% per year.

This ‘law’ was postulated by Jakob Nielsen of the Nielsen Norman Group in 1998 and subsequently updated in 2008 and 2019. Nielsen started by looking at usage for himself and other big data users, going back to a 300 bps (bits per second) modem used in 1984. In 1998 Nielsen had measured growth at 53% annually and rounded to 50%. In the ten years from 1998 to 2008, he had measured growth to be 49% annually. At least for himself and other big data users, this ‘law’ has held steady for 36 years.

While this is not really a law, but rather an interesting observation, it’s something that all ISPs should notice. In my time in the industry, I’ve seen the bandwidth use of the largest users grow at a faster pace than everybody else. This is something every network engineer ought to keep in mind when designing networks.

Consider bandwidth at schools. I recall seeing some schools got gigabit connections a decade ago. When first installed, the gigabit connections seemed to be oversized, and schools wondered at the time if they needed that much bandwidth. But since then, they’ve figured it out, and many schools have grown past a gigabit connection and want a lot more. School networks that were thrilled to find an ISP that could provide a gigabit of bandwidth are now looking to build private fiber networks as the most affordable solution for satisfying the bigger bandwidth needs they see coming in the future years.

We’ve seen the same thing at hospitals, factories, and other large businesses that have embraced the cloud. Businesses subscribe to large data pipes and then outgrow them in only a few years.

Network engineers are generally cautious people because they have to balance capital budgets against future broadband demand. Given an unlimited budget, many network designers would oversize data pipes, but they don’t like to be accused of wasting money. I can’t even count the number of times I’ve heard from network engineers who thought they were designing a network ready for the next decade only to find it full in half that time.

Nielsen points out a statistic that most of us have a hard time grasping. A network experiencing 50% annual growth will use over 57 times more data a decade from now than used today. He compares the growth of bandwidth to Moore’s law that says that computer chip capacity doubles every 18 months. That works out to mean over 100 times from capacity after a decade of growth.

The only place we see this kind of rampant growth for entire networks today is urban cell sites where data usage is doubling every two years. That’s a startling growth rate when you think of it in real terms. A cellular carrier that finds a way to double the capacity of an urban cellular network will see that new capacity gobbled up within two years.

It’s not hard to understand why the cellular industry is in a panic and is looking at every way possible to expand capacity. Interestingly the industry elected to hide their concern about growth behind the story that we need to do everything possible to enable 5G. I guess it’s hard for the cellular industry to expose their vulnerabilities by instead just telling the public that they need to expand cellular capacity greatly. This need for capacity is why they are building small cell sites, buying more spectrum, and pushing their labs to finish the development of 5G—they need all three of those things just to keep up with growing demand.

All network owners need to acknowledge that there are parts of their network where the demand is growing faster than average, and any network updates should make certain that the largest customers get the future capacity they are sure to need.