You Can Never Have Too Much Bandwidth

Internet Spotlight

By Jayne MillerSep 29, 2017


When our VP of Strategy Stephan Beckert agreed to sit in the hot seat for this month's Spotlight, I knew I wanted to ask him about the state of the global internet.

Where are new cables? Who is investing in them? What do they mean? The internet is growing, right? How are content providers contributing to this story?

I guess it's safe to say that this month's Spotlight is all about the big questions.

Stephan not only tackled my web of interconnected internet queries, but he painted a picture about how the internet we know and love is growing in 2017, and why that growth looks different than it did even just five years ago.

You can read our conversation below or listen to the whole thing here.

Jayne Miller: Thank you for taking the time, Stephan.

Stephan Beckert: Thank you for having me.

JM: I’m going to dive in with a big question. And that big question is: who is investing in submarine cables? What is happening?

SB: It’s been interesting to see how the sub cable market has evolved over the years. There was the crazy period in the early 2000s followed by sort of a nuclear winter. Now we’re in more of a stable period. Since about 2008 we’ve seen steady annual investment in the range of one-half to $2 billion annually.

And over the last few years we’ve seen a noticeable increase in the share of investment that’s coming from consortium systems as opposed to privately-funded systems.

By consortium systems, we mean cables that are funded by a group of carriers or content providers with or bandwidth requirements of their own. So we’re talking about large content providers like Google, Facebook, Microsoft, and-to a lesser extent—Amazon and others. And also large ISPs (internet service providers) who are investing in a wide range of systems.

Since about 2008 we’ve seen steady annual investment in the range of one-half to $2 billion annually.

For example, the INDIGO-West cable system in Australia where some of the primary investors include Telstra, Singtel, and also AARNet and Google. So you’ve got a content provider, you’ve got a research and education network, and two incumbent telcos all investing in a cable system. Most of the investment is coming from organizations that need extra capacity on that route, not because their business plan involves selling a lot of capacity on that route.

JM: I hear you mentioning Microsoft, Amazon, and Google and I’m hearing “content providers.” So maybe we could talk a little bit about their story.

SB: What we’ve seen in the last few years—there’s been a shift in the nature of what’s been driving international bandwidth requirements. Starting sometime in the second half of the 1990s until probably about five or six years ago the primary driver of international bandwidth requirements have been ISPs who need international capacity to connect their end users to upstream service providers and vice versa.

That was, for much of the last 15-20 years, the primary driver of international network requirements. They accounted for roughly 80 percent or so of used international capacity.

The other main driver of international network usage that we track are private networks—content provider’s networks as well as networks of governments and corporations and the research networks of groups like AARNet or GÉANT.

And we saw that the growth rates of private networks and internet backbones tended to parallel one another fairly well. Not always at the same pace on an annual basis, but compounded they grew at roughly a similar pace for the better part of a decade. Private networks’ share stayed at around 20 percent while ISPs stayed at around 80 percent.

But starting around 2012 we saw those growth rates starting to diverge.


Since 2012 the international bandwidth requirements of content providers have consistently outpaced those of international ISPs. So we’ve seen content provider’s network requirements growing at about just under 70 percent annually over the past five years compounded. While the international capacity requirements of ISPs have grown more in the range of 30-35 percent annually. Which is still fast by the standards of any rational industry. But it’s also down from the very high growth rates we saw in the early years.

Since 2012 the international bandwidth requirements of content providers have consistently outpaced those of international ISPs.

In 2016, the ISPs accounted for about 54 percent or so of used bandwidth, while the share accounted for by private networks—content providers in particular—has grown to roughly about 46 percent. We’ve seen a real shift.

JM: We’re not saying the internet has stopped growing [because of this shift]. It is a change in how fast.

SB: Yeah and when we look at growth rates it’s important to remember that these growth rates reflect the use of caching technologies and many other techniques used to dampen aggregate bandwidth demand and to localize content. Even in the slowest growth markets—like the U.S. and Europe—we’re still seeing annual growth rates in the range of 30 percent or so annually. I think it’s fair to say you can never have too much bandwidth.

JM: The other question I had was something we were talking about earlier. The implications of content being closer to the end user, because that’s what’s happening, ultimately?

SB: Right. And the reason content providers are investing in cable systems is because they are pushing large data centers ever closer to large end user populations. The (first) major data center deployments were obviously North America. They were followed by large deployments in Europe and now, increasingly, Asia. There are a few in Latin America and other regions, but really the dominant places for those data centers are North America, Europe, and Asia.

The reason content providers are investing in cable systems is because they are pushing large data centers ever closer to large end user populations.

They’re doing this for various reasons. For example, the performance of their services. Much faster response times and content delivery times. 

The other factor that plays into this sometimes is laws regarding customer data and where it can be stored. Data retention laws that require certain content about customers to be stored in specific ways and locations. That factors in particularly to corporate cloud services. But since many of these companies are also the leading providers of corporate cloud services, they tend to go hand in hand.

JM: Any predictions?

SB: Well, there’s one element that I think has been a bit under appreciated. It’s kind of interesting to think about that 10-15 years ago there was a lot of discussion in international internet governance and policy bodies—for example, the OECD, the United Nations, ITU, and so on—a lot of discussion about a sort of colonial pattern of international internet traffic. And sort of an implicit surcharge on internet usage by developing countries. Charges that ISPs in developing countries were subsidizing. ISPs in North America, in particular.

These overseas internet service providers had to lease submarine cable capacity to the U.S. to connect to upstream transit and content providers. And by doing this they were implicitly subsidizing the U.S. ISPs who could connect to the overseas ISPs without having to incur those same kinds of investments.

And what’s happened now in the last five to six years is that these content providers have become the dominant users of bandwidth on a couple of major routes and are bringing that content ever closer to end users. And as a result we’re seeing service providers able to connect to ever more content locally, and to peer ever more of their traffic locally.

We’re seeing service providers able to connect to ever more content locally, and to peer ever more of their traffic locally.

For example, I spoke recently with an Asian ISP that reported they’re now able to peer about 90 percent of their traffic locally. As a result their international internet bandwidth requirements have been almost flat for the last few years because the share of traffic they’re connecting locally has grown so fast relative to the share of traffic they need to route overseas.

So what’s happened is that we’re seeing—by making these investments in cable infrastructure—the content proprietors are actually reducing the costs that retail ISPs incur to deliver content to their end users.

The flip side of that same coin is that as a result of these content provider investments in network capacity and data centers, we’re seeing demand growth for long-haul network services by ISPs slowing. While this may reduce costs for the retail division of the same overseas telco, it could also reduce the potential revenue opportunities for their wholesale division.


New call-to-action

Stephan Beckert

Stephan Beckert

Stephan was formerly TeleGeography's VP of Strategy. He was responsible for new product development and advised TeleGeography's research teams.

Connect with Stephan